Connected Speech and Syntactic Impairment in Primary Progressive Aphasia Describing the diagnosis of Primary Progressive Aphasia and its variants with particular focus on the speech and language features that differentiate them. Presentation Video
Presentation Video  |   November 01, 2015
Connected Speech and Syntactic Impairment in Primary Progressive Aphasia
Author Affiliations & Notes
  • Naida Graham
    University of Toronto
  • Presented at the 25th Annual Research Symposium at the ASHA Convention (November 2015).
    Presented at the 25th Annual Research Symposium at the ASHA Convention (November 2015). ×
  • The Research Symposium is hosted by the American Speech-Language-Hearing Association, and is supported in part by grant R13DC003383 from the National Institute on Deafness and Other Communication Disorders (NIDCD) of the National Institutes of Health (NIH).
    The Research Symposium is hosted by the American Speech-Language-Hearing Association, and is supported in part by grant R13DC003383 from the National Institute on Deafness and Other Communication Disorders (NIDCD) of the National Institutes of Health (NIH).×
Article Information
Speech, Voice & Prosodic Disorders / Apraxia of Speech & Childhood Apraxia of Speech / Special Populations / Older Adults & Aging / Research Issues, Methods & Evidence-Based Practice / Language Disorders / Aphasia / Attention, Memory & Executive Functions / Clinical Practice Research / Diagnostics, Screening and Assessment Research
Presentation Video   |   November 01, 2015
Connected Speech and Syntactic Impairment in Primary Progressive Aphasia
CREd Library, November 2015, doi:10.1044/cred-pvd-c15004
CREd Library, November 2015, doi:10.1044/cred-pvd-c15004

The following is a transcript of the presentation video, edited for clarity. Click the PDF icon to download the presentation slides.

I'll be talking today about connected speech and syntactic impairment in primary progressive aphasia. This is my disclosure, I'm employed at University of Toronto and the research was funded by the Canadian Institutes of Health Research.
So in this talk, I'm going to start off by describing diagnosis of PPA and its variants. We've actually already heard most of this twice already today, but I will go over it again just because some people have just arrived for this session. Then I'll describe connected speech in each variant with a particular focus on syntactic production, and then I'll describe three studies that I was involved with where we evaluated connected speech in PPA, and finally I'll finish with conclusions.
Diagnosis of PPA
So starting with diagnosis of PPA. I'll just start with a definition actually, so PPA primary progressive aphasia is a dementia in which language is the earliest and most severely affected aspects of cognitive functioning. A large international group of experts in PPA established consensus guidelines for the diagnosis, and they were published in the article that I've got the reference for here, the article is by Gorno-Tempini et al. It was published in 2011 in Neurology.
To be diagnosed with PPA, the patient has to have has to meet basic PPA criteria. So first of all they should have insidious onset and gradual progression of language impairment, and the aphasia should at least initially, be the most salient impairment and should be the main factor contributing to disruption of activities of daily living. So this point really is saying the aphasia is primary, and finally diagnostic testing should suggest a neurodegenerative process. These are the inclusion criteria but basically to meet these criteria patients have to have a language impairment that is progressive and circumscribed.
These are the exclusion criteria for diagnosis of PPA. It should not be the case that the pattern of deficits is better accounted for by another disorder, like say another neurodegenerative disorder or vascular disease, and also there should not be prominent initial cognitive impairments outside the language domain. There also should not at least initially be a prominent behavioral disturbance, involving for example disinhibition compulsive behaviors or personality change.
So I've just described what's involved in making a root diagnosis of PPA, and then clinical diagnosis of the variant can occur when a case has a speech and language features that are characteristic of that variant.
So now I'm going to describe those features starting with the semantic variant. So in this variant there are impairments in naming and impairments in single word comprehension. These are core features and both are essential for diagnosis. You can see impairments in naming really in any of the PPA variants but the impairment is usually more severe in the semantic variant. And then to meet diagnostic criteria for this variant, the patient also has to have at least three of the four subsidiary features and these are spared repetition, spared speech production with respect to grammar and motor speech, impaired object knowledge, and surface dyslexia or dysgraphia.
To be diagnosed with the non-fluent variant a patient has to have at least one of the core features. The first is agrammatism in language production. And the term agrammatism has been used inconsistently in the literature, and this is something I'll be returning to shortly. The other core feature of the non-fluent variant is effortful halting speech with distortions and inconsistent speech sound errors, in other words apraxia of speech. So to be diagnosed with this variant a patient has to have either agrammatism in language production, or apraxia of speech or often they have both of these features. And then they would also have to have at least two of the three subsidiary features to meet criteria for the non-fluent variant, so these are first of all impaired syntactic comprehension, spared single word comprehension, and spared object knowledge.
The core features of the logopenic variant are impaired single word retrieval and impaired repetition of sentences and phrases. Impairment in word retrieval can also be observed in this semantic variant but usually the word finding impairment is more severe in the semantic variant than in the logopenic variant. And then to be diagnosed with the logopenic variant, a patient also has to have at least three of the four subsidiary features. So these are spared motor speech, absence of frank agrammatism, spared single word comprehension and object knowledge, and production of phonological errors.
Based on the Gorno-Tempini et al. criteria, these are the areas of predominant atrophy in each PPA variant. In the semantic variant, the abnormality is in the anterior temporal lobes and it's usually greater on the left than the right as is the case as in this illustration here. In the non-fluent variant, the predominant atrophy is in the left posterior frontal insular region, and in the logopenic variant predominant atrophy is in the left posterior perisylvian or parietal region.
Description of Connected Speech in Each Variant
So now I'm going to describe connected speech in each variant, and I'm going to focus particularly on quantitative analyses that have been published in the literature.
Starting with the semantic variant it's been shown that these patients have normal rates of syntactic errors, phonological errors, and false starts field pauses, and repaired sequences.
On the other hand, relative to controls, they have reduced proportions of open class words, they have an increased reliance on general terms, and they use words that are higher in frequency and familiarity, as well semantic variant patients use more pronouns and more pronouns with ambiguous reference. So all of these features arguably arise from the semantic impairment.
So now focusing on syntactic skills in the semantic variant, it's been found that rates of syntactic errors are no higher than controls, but it's been found that the mean length of utterance is reduced and that speech is reduced in syntactic complexity relative to controls. Also semantic variant patients rely on a restricted range of syntactic constructions. So altogether this suggests that there may be a mild expressive syntactic impairment in the semantic variant.
Okay, focusing now on the non-fluent variant, it's been found that relative to controls, these patients produce fewer words, they use shorter utterances, and they speak more slowly. As well they produce more phonemic errors and more distortions.
Early clinical description suggested that the language impairment in the non-fluent variant parallels that seen in non-fluent stroke aphasia. So the idea is that in non-fluent PPA, the patients have got something like a progressive broke as aphasia. Subsequent investigations have both supported and disputed this idea, but the idea that the aphasia is comparable in these distinct pathologies persists and indeed there are clear similarities.
Before I go on and describe the grammatical impairments in the non-fluent variant, it seems pertinent to give a definition of agrammatism since this is a core feature of the variant. A large body of work, based mainly upon patients with stroke aphasia, has identified the main features of agrammatism and in spoken language. So first of all there's a tendency to omit and or substitute grammatical morphemes and this leads to syntactic errors, as well there are reductions in syntactic complexity in proportion of verbs and in proportion of closed class words. Although agrammatism as I just described it is an established concept in aphasiology, the term is also sometimes used more loosely to mean any type of grammatical impairment. So I'm going to use the term frank agrammatism to refer to the grammatical impairment that I've just described here.
Okay, so now focusing on syntactic skills in non-fluent PPA patients. Findings have shown that these patients produce fewer complex grammatical structures and controls, they produce fewer well-formed sentences, and they make more grammatical errors. But in some reports the presence of grammatical errors is a requirement for the diagnosis of the non-fluent variant, so then it's inevitable that these patients will be found to produce more grammatical errors than controls.
In contrast, some studies have found normal rates of grammatical errors in non-fluent PPA patients or they found that these errors are made only by a minority of patients.
Although frank agrammatism of the type seen in Broca's aphasia occurs in the non-fluent variant, there are findings suggesting that it's not universal. Group studies have shown for example that the speech of patients with the non-fluent variant may include normal proportions of verbs and normal proportions of closed class words.
So these are findings which you would not expect in frank agrammatism. Other studies of course have found contradictory result, Wilson et al found that their group of non-fluent PPA patients produced reduced proportions of verbs and reduced proportions of closed class words, but they noted that this significant group effect was actually driven by a minority of patients, and that most of their non-fluent patients produce normal proportions of verbs in closed class words.
So this review of the literature shows that features of frank agrammtism are inconsistently documented in group studies of non-fluent PPA. When features of agrammatism are documented there may be individuals whose results do not follow the group pattern. So all of this suggests that across patients, there's a high degree of variability in grammatical skills.
Turning now to the logopenic variant, because this variant was identified more recently than the other variants, there are fewer studies which characterized speech in this area but there are still some. So overall, quantitative analyses have shown that the speech of logopenic patients shows numerous signs of dysfluency. So their speech rate is lower than controls and logopenic patients make a high rate of pauses. They also make a lot of false starts and hesitations and make a lot of filled pauses, and finally they have an increased number of repaired sequences.
The speech of logopenic patients also shows signs of word-finding impairments, so I mentioned that they have an increased pausing rate, the pauses have been found to occur especially before nouns. And also logopenic patients use an increased number of pronouns and a reduced proportion of closed class, of open class words. At the single word level, phonological errors are made by some patients but speech is usually well articulated and with no distortions.
Focusing now on syntactic production in the logopenic variant, studies have shown that these patients produce a, have a mean, reduced mean length of utterance relative to controls and also that the proportion of grammatical sentences is lower than controls. It has been found though that logopenic patients produce a higher proportion of grammatical sentences than the non-fluent variant patients. Also in logopenic patients, it's been found that they make an increased rate of syntactic errors, this was found by Wilson et al but they attributed this to the constant rewordings that these patients do.
So all of this leads to the question "Is there a syntactic impairment in production in the logopenic variant?" And I think across the literature different people would give different answers to this question. So Tyson et al concluded that yes there is a syntactic impairment, but that it's mild and occurs in a minority of patients. On the other hand Thompson et al concluded that no there is not a syntactic impairment in logopenic PPA, but it's worth noting that part of the way they diagnose the logopenic variant was that patients had to pass a syntactic production test.
Study 1: Automated Analyses of Connected Speech in PPA
Okay, the studies that I've referred to so far have all used manual counts of features from transcriptions of speech in PPA patients. I'm going to tell you now about a study that colleagues and I did, where we did automated analyses of connected speech in PPA. In this study, we used software to analyze transcriptions of speech samples, and using this method we were able to analyze a larger range of features that have been addressed in many other studies. The task that we used to generate the narrative speech was recounting of the Cinderella story. The participants were ten people with semantic variance and fourteen people with non-fluent variant PPA, and there were sixteen controls. We didn't have enough logopenic patients to include them in the group analysis. Using computational techniques, syntactic and semantic features were automatically extracted from transcriptions of narrative speech.
So this slide and the next slide give you some examples of the automatically extracted features. There were twelve part of speech features, so we had counts of number of nouns, verbs, adjectives and pronouns for example. The part of speech tagging was done automatically by computer, but we also had a human annotator perform the part of speech tagging for about a quarter of the narratives, and we found that the agreement between the human annotator and the computer annotation was about 90%. The automatically extracted features also included 26 complexity features, for example mean length of sentence and mean length of clause, and also there were counts of the number of coordinate conjunctions.
There were five fluency features. These included; word lengths in terms of number of letters, and also the total number of words produced in the narrative, as well as speech rate in words per minute and production of fillers like am and ah. And finally there were eleven psycholinguistic features, so these included its frequency familiarity and age of acquisition for the words produced, and then this was also analyzed specifically just for the nouns and just for the verbs, and we also had counts light verbs and heavy verbs.
We used partial least-squares analysis to identify the features that best predicted group membership, and the features on this slide all distinguished well between the semantic variant patients and controls. So focusing on the left-hand box, the things that were elevated in the semantic variant patients were first of all frequency and familiarity, so they used words that overall were higher in frequency and familiarity, and this was especially true of nouns. So it was a separate finding that they also used nouns that were higher in frequency and familiarity than controls. The semantic variant patients used the higher number of demonstratives and controls, and the demonstratives were words like these, those, this, that, here, there. This is consistent with other findings in the literature suggesting that semantic variant patients use less specific words, but we think that this is the first time this is been specifically examined. The semantic variant patients also produced more clauses than the controls. The and yea, if you look on the right-hand box the clauses were overall of shorter length. The explanation for this is not entirely clear but we think it might be due to the fact that the patients had a tendency to produce a lot of fillers like "you know", and "what do you call it", and these are relatively short pauses which also increased to, in the, increase the account for the number of clauses.
Now focusing on the right-hand side of the slide the things that were reduced in semantic variant PPA, they produced overall proportionally fewer nouns than the controls did, and also the nouns-to-verb ratio was lower in the patient's. Difficulty, this difficulty with production of nouns is an expected finding, but unexpectedly we found that the patients with semantic variant PPA used words that were shorter in length than the words used by the controls. The effect was small but it was significant. We attribute this to availability of words rather than difficulty with pronouncing words, because the controls were more likely to use longer words like Cinderella and beautiful, they used these words more often than the patients and this may explain the finding. But overall the features that distinguished the semantic variant patients and controls can largely be attributed to the semantic memory impairment in the semantic variant patients.
This slide shows the features that best distinguished the non-fluent PPA patients and controls. So the non-fluent patients used words that were higher in frequency overall than the words used by controls, but this is, as a separate finding, we found that the non-fluent patients use verbs that were higher in frequency than the, the verbs used by controls. I just wanted to mention that this finding is not due to excessive reliance on light verbs like be, do, go etc. all of which are very high frequency word, very high frequency words but we analyzed this and found that there was no difference between the non-fluent patients and controls with respect to proportion of light verbs. And also it's worth noting that the non-fluent patients produced nouns and verbs, well verbs in normal proportions because the groups didn't differ on noun-verb ratio. Turning now to the things that were reduced in the non-fluent variant relative to controls, the controls not surprisingly had a short, lower speech sorry, the non-fluent patients had a lower speech rate than the controls and this is been documented in many other studies. In our study the lower speech rate was the feature that best distinguished the non-fluent patients from controls. We also found that the non-fluent patient's overall use shorter words than controls. This fits with findings from earlier studies that show that increased word length has a negative effect on naming and reading and repetition in non-fluent PPA patients.
So just to summarize the findings, relative to controls the semantic variance and non-fluent PPA patients used words that were higher in frequency, especially nouns for the semantic patients and verbs for the non-fluent patients. Relative to controls the semantic variant patients used words which were higher in familiarity, especially nouns, and they produced fewer nouns but more demonstratives. The non-fluent patients had slower speech than controls and used shorter words.
Surprisingly none of the grammatical features distinguished the non-fluent PPA patients from the controls and this include, these features included measures like sentence length, clauses per sentence, and other measures of other metrics of syntactic complexity.
Study 2: Evaluation of Agrammatism in Nonfluent Variant PPA
So now I'm going to go on and tell you about another study where we evaluated specifically agrammatism in the non-fluent variant. There are at least two potential reasons for the variability in grammatical skills that's seen in the literature on non-fluent PPA as I've just highlighted. So it could be first of all that studies legitimately include variable numbers of patients whose dysfluency arises at least partially or even primarily from a motor speech impairment. Some researchers separate patients whose non-fluent progressive aphasia is associated primarily with the grammatical or primarily with the motor speech impairment but others do not. The other potential reason for this variability in findings with respect to grammatical skills is that it could arise from difficulty with distinguishing the non-fluent and logopenic variants. So this can happen in two ways. First of all it could be that logopenic, logopenic patients have been inadvertently included in groups of non-fluent PPA patients, and then this would make the non-fluent group seem less grammatically impaired. The inclusion of logopenic patients in non-fluent groups is particularly an issue for studies published prior to 2011, when the diagnostic criteria for the logopenic variant were first published. So the flipside of this is that it could be that patients with non-fluent PPA are classified as logopenic because they don't have frank agrammatism, and indeed some people in the literature say that they have a policy of doing this. This would then mean that the patients classified as non-fluent, the ones that are still left in the non-fluent group would be more grammatically impaired.
We tried to address both of these issues in the study I'm about to tell you about, so we addressed the issue about variable numbers of patients with a motor speech impairment by doing a thorough evaluation of motor speech skills in our PPA patients, and then in this way we were able to look at grammatical production in a set of patients for whom motor speech impairment has been ruled out as a contributing cause of the dysfluency. We addressed the issue about distinguishing non-fluent and logopenic PPA patients by using volumetric MRI data to provide unbiased imaging support for the diagnosis and then in this way we were able to look at grammatical production and a set of patients for whom the clinical diagnosis of the non-fluent variant is unequivocally imaging supporting.
So now I'll go over the specifics of the study, so the study was motivated by the inconsistent results in the literature with respect to grammatical skills in the non-fluent variant. We aimed to examine syntactic production and non-fluent PPA patients, who had both preserved motor speech skills and whose diagnosis was independently supported by volumetric MRI data. The participants were fourteen patients with non-fluent variant and fourteen semantic variant PPA patients, and there were four logopenic variant patients.
Blinded expert raters evaluated speech samples for features of agrammatism and apraxia of speech. The raters were blinded to the variant that is, of PPA that a patient had been diagnosed with, and this was the reason that we included patients with all three variants in this study, although the study is primarily about the non-fluent patients. So the Raters were blinded as I say to the diagnosis of each individual patient, but they also didn't know how many patients they were of each type. There were two raters, they're both speech language pathologists. They rated the recordings that they listened to or watched separately and the agreement between them was high. So the, they watched the video recordings of narratives and the narratives comprised topic directed interviews. These are interviews in which patients are asked to describe a series, of standard series of topics including for example; they're asked tell me about what you do each day, tell me about your family, tell me about your health right now. The Raters also listened to audio recordings of tasks that are known to be sensitive to apraxia of speech. So these tasks were, were repetition of words of increasing length, repetition of polysyllabic words and phrases three times and rapid repetition of pataca. The Raters used checklists to look for features of agrammatism, and features of apraxia of speech.
This slide gives some example features from the agrammatism checklist, so these features included lack of functor or closed class words, omission or substitution of inflectional affixes, and sentences that are generally simple and incomplete, and limited variety of sentence structure.
This slide gives some example features from the apraxia of speech checklists so the Raters looked for syllable segregation, phonemic anticipatory perseverative or transposition errors, intrusion of schwa and articulatory grouping.
So volumetric MRI data were analyzed in this effort to provide independent support for the clinical diagnosis of the variant. We analyzed regions of interest, all of which were on the left side of the brain, and specifically we analyzed regions that are expected to be atrophic in any variant of PPA. Atrophy scores were calculated for each individual for each brain region, and these atrophy scores took account of variation in brain or head size.
This slide shows the criteria that we used to decide whether there was imaging supported diagnosis of the, of the non-fluent variant, so atrophy scores had to be both abnormal in the insula and/or inferior frontal regions, which is what you would expect in non-fluent PPA, and atrophy scores had to be normal in the inferior parietal and posterior temporal regions. The idea with the second criterion is that it would rule out the possibility that any patients in the non-fluent group may have had the logopenic variant. This is actually called the quite a stringent criterion because if you look in the literature you see that many non-fluent PPA patients have had spread of their pathology more posteriorly into the inferior parietal and posterior temporal regions.
Okay, so turning to the results, I'm going to just focus on the non-fluent patients. So three of them were found to have frank agrammatism and apraxia of speech, and one of them had apraxia of speech only. Ten of the patients had no frank agrammatism or apraxia of speech.
For five of these patients, the volume in, the volumetric imaging data unequivocally supported the diagnosis of the non-fluent variant, so it seems unlikely that these patients had logopenic PPA even though they had no frank agrammatism or apraxia of speech. For further five patients, they're in the bottom right of this slide, these patients showed the same clinical syndrome and they had left inferior frontal atrophy, but in these cases there were abnormal atrophy scores in the left posterior temporal and/or inferior parietal regions. So although this is not necessarily incompatible with the diagnosis of the non-fluent variant, it didn't allow us to rule out the possibility based on imaging that these patients have the logopenic variant. So you may be wondering what speech is like in these patients who are non-fluent but without frank agrammatism.
This is a transcription of a patient who is, was judged by our raters not to have frank agrammatism or apraxia of speech. The patient is responding to the request to tell me about what you do each day, and this transcription is only part of their response. I'll read the response I can't reenact the disfluency, but you'll at least be able to see the grammatical structure of the speech. So the patient said, "I read a lot, and I look at television here and there, and then I do go out. I don't do very much really, I go visiting and have an odd person in to have a cup of tea and then on the weekends I go out with my family".
The diagnostic criteria published by Gorno-Timpini et al do acknowledge that lack of frank agrammatism or apraxia of speech can arise in the non-fluent variant, at least in the early stages, and so therefore we looked at symptom duration in the, in our patients -- the ones who had neither frank agrammatism nor apraxia of speech.
So we found that the mean symptom duration was 3 1/2 years, but three of the patients had a history that was longer than four years, in fact for two of them it was just over six years. So it seems that some of these patients were beyond the earliest stages of illness.
The main findings from this study is that some of the non-fluent PPA patients had a clearly imaging supported diagnosis, but did not have frank agrammatism in speech or apraxia of speech.
The implications of this study are first of all for differentiation of the non-fluent and logopenic variance. Some researchers classify as logopenic patients whose speech is non-fluent, but without frank agrammatism or apraxia of speech. The present results suggest that this would lead to misclassification in some cases. We also noted that agrammatism in production in the non-fluent variant may be difficult to detect because in some patients it may be subtle, and this can be the case even beyond the earliest stages of illness.
Study 3: Agrammatism in Spoken vs. Written Production in nfPPA
Okay, so now I'm going to just briefly describe a study that I did, done prior to the other two, but this study focused on agrammatism and spoken verses written production in the non-fluent variant. In this study there were fourteen non-fluent PPA patients and eleven controls. The study was done prior to the publication of the current diagnostic criteria, this study was published in 2004. So I've looked retrospectively at the information I have on these fourteen non-fluent PPA patients and I think there's evidence that at least twelve of them are unlikely to have had the logopenic variant. So nine of them had imaging with MRI and/or PET which showed that the predominant abnormality was in the left frontal insular region and four of these twelve patients develop cortical basal syndrome, which is something that's much more closely associated with non-fluent PPA then with the logopenic variant. For the remaining two patients in the non-fluent group, I just didn't have enough information to be able to distinguish whether they had the non-fluent or PPA variants. There was nothing to indicate that, sorry, whether they had a non-fluent or logopenic variants. There was nothing to indicate particularly that they were logopenic, but I just didn't have enough evidence to address this.
In this study, we analyzed written and spoken descriptions of the cookie theft picture. So I'm not to go into too much detail about all of the results, but I just wanted to highlight a couple of things. So first of all the non-fluent PPA patients had normal noun-verb ratios, and normal content-to-function word ratios, and normal rates of syntactic errors. So they didn't exhibit features of frank agrammatism. One thing I wanted to highlight is this is an entirely different set of patients. There's no overlap with the patients in the studies I just described and these patients also are diagnosed by a different set of neurologists. We also found, we had findings with respect to written picture description, so we have the same results that overall the groups produced the non-fluent group produced normal noun-verb ratios and normal content-to-function word ratios, but the patients produced more syntactic errors than controls in the written picture description. So basically we found that some patients make syntactic errors in written production when they don't make them in spoken production.
So once again meaning, once again in another study, the results suggest that some non-fluid PPA patients don't have frank agrammatism. We also found that agrammatism may be apparent in writing before it's noticeable in spoken language.
Final Comment & Conclusions
So I'm going to now just finish with one slide that's got some final comments, and then I'll finish the conclusions. Now, because I've noted that there's variation in the literature with respect to how PPA patients are subtyped, it seems that we need better ways to systematically evaluate grammatical production. So most of us evaluate grammatical production by listening very carefully to patient speech, and I'm certainly not saying that we shouldn't do that, but I looked for some more systematic methods of evaluating grammatical production.
So I've identified some assessments in the literature that may be useful. So first of all there's the make a sentence test which is by Billette et al in their publication in Aphasiology they've included all of the stimuli for this test, so in this test patient is shown written words -- so for example Emma baked pie party-- and then the patient has to produce a spoken sentence and inflect the words. So then they correct answer here would be "Emma baked a pie for the party."
Results from analyses done by Billette et al show that performance on their make a sentence test was strongly correlated with measures of patient spontaneous speech, especially production of grammatical errors and abandoned utterances. Billette et al compared performance on their test with performance on the Northwestern anagram test, which is by Weintraub et al. So that's another test in which patients have to make sentences, but they just have to arrange words to make a sentence, they don't have to speak the sentences in that test. Billette et al found that the make a sentence test was less prone to ceiling effects and therefore was better at picking up subtle early deficits.
There are also two rating scales that have been developed specifically to evaluate language impairments in primary progressive aphasia, and both of these scales provide ratings for severity of impairment in grammatical production and morphology.
So now just finishing with conclusions, in the semantic variant there's evidence that grammatical production is simplified although without errors. In the non-fluent variant there's a high degree of variability across patients with respect to grammatical skills, and in the logopenic variant, some patients may exhibit grammatical dysfunction but the impairment is usually mild and occurs in a minority of patients. Further research is required to clarify the differences in connected speech of patients with the logopenic variant in comparison with patients who have the non-fluent variant but without frank agrammatism or apraxia of speech.
And I want to acknowledge the contribution of my collaborators on the studies that I just described. Thank you for listening and I'd be happy to take questions.
Questions and Discussion

Audience Question

Hi, I'm Alyssa Lanzi from Duquesne University. I was wondering how a patient's cognitive decline can affect their discourse measures that you've seen in your research.

Well, the, of course for the, the PPA patients, they at least initially their, their cognitive decline should be just in language. So really I've just described what happens with a decline in language. There are other studies that look at connected speech in Alzheimer's for example, and as I recall those studies show that the, their speeches are lower in proportion of verbs and syntactically simpler. So there are some similarities with the results I've described from PPA patients.

Audience Question

I just would follow up. I think Alyssa's question is a good one because perhaps we could turn it around and say "Do you think in the language domain there are some early indicators that widespread cognitive decline is about to, or is beginning to happen?

So that's a good question and I think some of the indicators are, so I'm not basing this on data, more so on and you know, having done these analyses and been familiar with their speech, but I think some of the indicators that you know, this incipient or beginning cognitive decline beyond language, is that the patients don't follow the instructions as well. So they don't stay on the topic. They have a lot more comments that are off-topic like you know, the on the cookie theft picture, I do know they say, you know, "we have I like cookies" or you know, things that are just kind of irrelevant but, you know, so they they seem to talk, talk off topic more. Also of course the the word finding difficulty gets worse too, and that certainly affects speech I think really in any of the variance.

Audience Question

I've been wanting to ask someone like you just for a couple of days because now I've listened to about six hours on primary progressive aphasia. A prevalent symptom -- language problem -- in aphasia with stroke is perseveration, and I haven't heard anybody talk about perseveration in primary progressive aphasia, and looking in the end, what you showed today of course when to set out, you looked at a heck of a lot of cookie thefts and so forth. So can you make any observations on that?

So it's an interesting question and the, I've certainly seen perseveration, but I feel like when it gets, there are a minority of patients that do this and they, but sometimes it can be really severe. I think back to a patient that I used to see who had semantic variant who was very a gnomic and as her language impairment progressed the number of words that she could use became very reduced because she was so overtaken by perseveration. So she was you know and occasionally, you know, the words that she was perseverating on would change, but, you know I remember, you know, couple of sessions, almost the only thing she could say was "sugar", and every time she, you know, went to try to name something, you know, she says "sugar", you know, "not sugar", you know, and then she just couldn't seem to stop herself from saying "sugar", and she even perseverated it in writing. So we'd be doing, you know, spelling of words to dictation, and she would perseverate on letters and, but I feel like this strong perseveration comes more when the patients either are getting are other frontal impairments, or at least later in the course of the illness.
Audience Comment:
That's a great question. It'd be great to be able to capture that and perhaps some imaging at some point to know where that perseveration spot is.
Audience Comment:
Nobody's ever found the perseveration spot. So I'm not sure if you'll find it's just frontal, because we see perseveration and there are several studies that have looked at it through all types of aphasia caused by stroke, with different lesion locations. So I'm, I'm surprised nobody has looked up perseveration on primary progressive aphasia.
Naida Graham:
Do you know though, it's hard to look at perseveration. So I've actually given some thought to this, years ago actually not recently, but you know, with some of the patients that I have seen that did a lot of perseveration and it's, it's a hard thing to measure because it you know it's inconsistent, different things set it off, you know they do it sometimes and not others, and it was just I couldn't really come up with a good study that was really going to evaluate perseveration, but I think it's a really interesting thing I wish we understood it better.

Audience Question

Were there any medication implications regarding your clients?

So the most of our patients are not being treated. So there just because there is, this no cure for progressive aphasia. So you know some of them are on various medications, but it's actually something that we deliberately adopted a policy of ignoring, or at least we could, we can't control for this in our research. It's not so much that we ignore it but we can't control for this because certainly for the research that were doing. We can't ask people to abstain from medications, it would be unethical. So for us it may be creating noise in our data, but it's just something we have to tolerate.
Something I found fascinating is to see all these MLU reduction in patients for progressive aphasia, but in terms of eliciting a response but when it comes to echolalia the syntax is perfectly intact. It is surprising. I guess everybody heard so the ideas in echolalia wise, it's surprising the syntax is completely intact, and it's surprising to me too with some of the more severe non-fluent patients that I've seen that who are was still producing syntactically correct utterances. It's you know, it's amazing how this holds up obviously some people are agrammatic you know, right from early on in the illness, but some of them even the people without motor speech impairment are still able to produce syntactically correct utterances, even when they've had the disease for a while and their speech is pretty non-fluent.