Awareness of deficits in Alzheimer's disease patients: Analysis of performance prediction discrepancies


Correspondence: Pascal Antoine, PhD, Université Lille 3, BP 60149, 59653 Villeneuve d'Ascq Cedex, France. Email:



Unawareness has been operationalized in terms of a discrepancy between the patient's self-reports and three main categories of standards: judgment of a relative, clinical assessment, and objective test performance. The purpose of this study was to develop a new measure of deficit unawareness based on multidimensional, isomorphic, simple tasks and to examine the relationship between this measure and neuropsychological tests.


Analysis was conducted on cognitive performance prediction discrepancies in a sample of Alzheimer's disease (AD) patients and a matched comparison group.


Patients rated their cognitive functioning more highly than their performance, but their overall self-reports were lower than the overall self-reports of the comparison group. AD patients performed significantly lower than their predicted scores in all Dementia Rating Scale (DRS) domains, in contrast to comparison participants, who did not consistently perform significantly lower across domains. All unawareness scores were moderately inter-correlated, except for memory, and all unawareness scores with the exception of memory were correlated with overall neuropsychological functioning.


A methodological and conceptual difficulty has been identified, and this raises the issue of the generalizability of studies with a focus on memory unawareness. The method proposed seems a good tool to assess the relationships between unawareness and several different aspects of cognitive functioning, in particular executive functioning.

UNAWARENESS IN DEMENTIA, or anosognosia, is an impaired ability to recognize the presence or appreciate the severity of deficits in sensory, perceptual, motor, affective, or cognitive functioning.[1] Such abilities are particularly important from a clinical point of view because effective care for patients with Alzheimer's disease (AD) and their relatives is dependent on an understanding of the extent of the unawareness displayed by the patient.[2]

Unawareness has been operationalized in terms of a discrepancy between the patient's self-reports and three main categories of standards.[2, 3] The most commonly used standard is the subjective judgment of a relative. Another widely represented standard is the clinical assessment of a professional caregiver. Anosognosia assessed in those manners may be constituted in part by individual factors relevant to the clinician or the caregiver. Those methods may also be affected by psychosocial factors relating to the context of medical investigation or relating to caregiver burden. Thus, it is preferable to compare patients' reports of their cognitive functioning with their own performance rather than with subjective ratings.[4] Another category of standard, less widely used in the literature, uses comparisons of patients' self-reports with their objective test performance.[4-7] This method reflects the cognitive formulation of anosognosia. This method, however, has inherent weaknesses and requires careful selection of measures in order to overcome potential conceptual and methodological difficulties. First, memory functioning has been the main focus,[2, 4] for which a large range of cognitive functions is preferable when the object is anosognosia assessment.[7] Second, self-predictions of cognitive functioning and performance on neuropsychological tests may not be closely related, giving spurious discrepancy values. Thus, isomorphic measures, for example with similar contents and scaling of responses, are needed.[8, 9] Third, it is difficult for patients to predict their performance on laboratory tasks or complex neuropsychological tests. The discrepancy with performance might reflect the participant's inability to accurately assess the nature of the test.[10] It would be preferable for methods to be based on familiar experiences or situations that are simple to appraise.

Given the heterogeneity and methodological weaknesses of the assessments of unawareness in AD, it appears that there are no straightforward relationships between deficit unawareness and neuropsychological functioning.[11] The relationship between deficit unawareness and performance on neuropsychological tests, however, is of interest in connection with neuropsychological explanations for anosognosia.[12]

The present study assessed unawareness from a cognitive perspective and to see whether a method using prediction performance discrepancies can help to better understand anosognosia. In particular, we studied: (i) the development of a new measure based on multidimensional, isomorphic, simple tasks; and (ii) the relationship between this measure and neuropsychological tests.


The sample consisted of 63 patients (21 men, 42 women) and a comparison group of 21 participants (seven men, 14 women) with similar mean age and similar sex-ratio. The comparison group consisted of participants without suspected dementia. Patients had a diagnosis of probable AD according to NINCDS-ADRDA criteria.[13] The severity of the disease was assessed using Mini-Mental State Examination (MMSE)[14] and the Dementia Rating Scale (DRS) scores.[15] The Dementia Rating Scale is a 36-task individually completed instrument. The DRS is relatively brief and easy to use, allowing for a low floor so that even severely impaired patients can be evaluated. The DRS is sensitive at the lower ends of functioning, and differentiates levels of deficits. Several different centers located in urban and rural areas in the north of France were contacted. The centers included those serving relatively autonomous elderly individuals (daytime centers, elderly residencies, retirement homes) as well as centers for dependent persons (medical-care homes). Mean age in the patient sample and the comparison group was 85.5 ± 7.6 years and 85.6 ± 8.7 years, respectively.

Consent was obtained from all of the participants prior to the testing session in accordance with the Declaration of Helsinki. Patient and comparison groups were assessed on an individual basis. The testing was conducted by research assistants who had an undergraduate degree in psychology and were specifically trained for this study.

Multidimensional Isomorphic Simple Awareness Assessment

The current study focuses on the development of a measuring tool of cognitive deficit unawareness in which subjective self-ratings are compared with actual objective neuropsychological test performances. This measure was termed the Multidimensional Isomorphic Simple (MIS) Awareness Assessment. Patients and comparison participants were asked to predict their DRS performance on a dichotomous scale for each of the tasks of the DRS after hearing and seeing the task and prior to actually performing the task. This procedure including the DRS as standard was conceived for three reasons. First, memory functioning is not the sole focus. The DRS investigates four other areas: Attention, Initiation-Perseveration, Construction, and Conceptualization, with respect to the multidimensionality principle. Second, the DRS is composed of simpler items than traditional cognitive tests, decreasing its susceptibility to floor effects. It is also easy for participants to predict their likely performance on such simple tasks, with respect to the simplicity principle. Third, self-predictions of DRS performance are closely related to performance itself, giving interpretable discrepancy values with respect to the isomorphism principle.

One week before the DRS, the MMSE was given before the prediction task: ‘You have just completed one test [MMSE]. Now another test like it has to be administered. I would like to know today how you estimate the difficulty level of this test: will this test be easy or hard for you? I will explain to you what you have to do for each question and you have to tell me whether you think that you will give a right or a wrong answer’.

In the DRS instructions, items are arranged hierarchically, so that more difficult items are presented first. If a patient performs adequately on the initial items within a given subscale, full credit is given for the remaining items in that same subscale. This procedure significantly shortens the total testing time for individuals with relatively intact cognitive functioning. Consequently, the prediction tasks are concerned only with the first items of each subscale. Each of those DRS difficult items and associated material were presented one by one by the researcher. A prediction score was calculated for each subscale and the overall DRS. There are four possible cases: when a participant predicts a correct performance, but does not perform adequately on the item, the predicted score is set as the lower correct performance and full credit is given for the remaining items in that same subscale. When a participant predicts a correct performance and performs adequately, the predicted score is set at the real performance level and full credit is given for the remaining items. When a participant predicts a wrong performance, but performs adequately on the item, the predicted score is set at the real performance level and full credit is given for the remaining items. When a participant predicts a wrong answer and does not perform adequately on the item, the predicted score is set at the real performance level for this item, as for the remaining items in that same subscale. This method enables the sub-assessments to be focused on because the difference between the actual performance and the predicted score is strictly due to the cases of wrong predictions of correct performance. Dementia patients typically exhibit underappreciation of the severity and extent of their impairments.[16, 17] Moreover, for each participant, the scores of unawareness are calculated by the simple subtraction between the real performance and the predicted score.

Comparisons were then made between the performance values using t-tests. In the patient group, we studied the measures using Pearson correlation coefficients to examine the relationship between unawareness and neuropsychological tests. In order to avoid inflated correlations due to confounded measures, however, several general unawareness scores have been calculated. For example, in order to investigate the correlation between attention cognitive functioning and the general unawareness score, the latter score has been calculated irrespective of the unawareness data regarding attention deficit. Thus, the correlation reflects the relationship between objective attention functioning and the unawareness level throughout cognitive dimensions (initiation, construction, conceptualization and memory), except for attention. The same method was used for each of the correlations. Along the same line of thought and so as to avoid inflated correlations, several global scores of cognitive performance have been calculated. For example, in order to investigate the correlation between attention deficit unawareness and the global score of cognitive functioning, the latter score has been calculated irrespective of the attention performance data.


Preliminary results

As shown in Table 1, MMSE scores of the patients and the comparison group were respectively 17.16 ± 3.96 (range, 11–24) and 27.95 ± 1.20 (range, 26–30). DRS scores were respectively 94.54 ± 19.46 (range, 39–122) and 136.19 ± 4.99 (range, 125–144; Fig. 1). As expected, a significant effect of group was found for cognitive level measured on the MMSE (t = 12.27, d.f. = 82, P < 0.001) and DRS (t = 9.67, d.f. = 82, P < 0.001).

Figure 1.

Differences between predicted and observed neuropsychological test scores for the Alzheimer's disease (AD) patients and comparison group. (a) Attention; (b) Initiation; (c) Construction; (d) Conceptualization; (e) Memory; (f) Overall. C_DRS, performance of the comparison group on the Dementia Rating Scale (DRS); C_Pred, prediction of the comparison group; P_DRS, performance of the AD patient group on the DRS; P_Pred, prediction of the AD patient group.

Table 1. Demographic and clinical data
 AD patients, n = 63Comparison group, n = 21 
n (%)n (%)t (d.f.)P
  1. AD, Alzheimer's disease; DRS, Dementia Rating Scale; MMSE, Mini-Mental State Examination.
Female42 (66.6)14 (66.6)  
Male21 (33.3)7 (33.3)  
 Mean ± SDMean ± SD  
Age (years)85.5 ± 7.685.6 ±
MMSE Score17.16 ± 3.9627.95 ± 1.2012.27 (82)<0.001
DRS Total Score94.54 ± 19.46136.19 ± 4.999.67 (82)<0.001

Comparisons of predictions with cognitive performances

As shown in Fig. 1, using a Bonferroni corrected t-test, comparison participants gave higher self-ratings than their objective performance for the subscore of attention (t = 4.7, d.f. = 20, P < 0.001) and for the overall DRS score (t = 3.4, d.f. = 20, P < 0.01), but the difference did not reach statistical significance for the other subscores. Significant differences, however, for the attention subscore and overall score were only weak, reaching 0.52 and 1.43, respectively (Fig. 1).

In contrast, in the patient group, all corrected t-tests reached significance (t > 6; d.f. = 62; P < 0.001). Overall DRS score and self-rating were respectively 94.54 ± 19.46 and 117.65 ± 11.68. Patients rated their cognitive functioning higher than their performance, but their overall self-reports were lower than the overall self-reports of the comparison group (t = 7.6, d.f. = 82, P < 0.001).

In the patient group, the predictions and performances had statistically significant positive correlations for attention (r = 0.75; P < 0.001), initiation (r = 0.41; P < 0.001), construction (r = 0.41; P < 0.01), conceptualization (r = 0.62; P < 0.001), memory (r = 0.37; P < 0.01), and overall DRS (r = 0.58; P < 0.001).

Intercorrelations between unawareness subscales in AD

As shown in Table 2, all unawareness scores were moderately intercorrelated except for memory. No significant correlation was found between the memory unawareness score and the unawareness scores for attention (r = 0.019; P = 0.14), construction (r = −0.02; P = 0.91), or conceptualization (r = 0.18; P = 0.15). Memory unawareness was significantly, positively related to initiation unawareness (r = 0.38; P < 0.01).

Table 2. Inter-correlations between scores of deficit unawareness
  Prediction performance discrepancy (unawareness)
  1. Pearson correlation coefficient; *P < 0.05; **P < 0.01.
Prediction performance discrepancy (unawareness)Initiation0.37**   

Cognitive correlates of deficit unawareness in AD

The next objective was twofold. In coherence with literature on the neuropsychological correlates of anosognosia,[18] we wished in the first place to study the relations between the general unawareness score and the objective performances for each cognitive function (Table 3, last line). Second, the aim was to study the relations between each of the different unawareness dimensions and the global level of cognitive alteration (Table 3, last column).

Table 3. Cognitive correlates of deficit unawareness
  Cognitive performance on the Dementia Rating Scale
AttentionInitiationConstructionConceptualizationMemoryTotal DRS
  1. Pearson correlation coefficient; *P < 0.05; **P < 0.01; ***P < 0.001. Italics, correlations between corrected scores. Overall scores were corrected by removing from their calculation the correlated subscales. For similar reasons, correlations on the diagonal were not calculated.
Prediction performance discrepancy (unawareness)Attention−0.37**−0.37**−0.36**−0.27*0.45***
General unawareness score0.34**0.55***0.41***0.44***0.42***

As shown in Table 3, the general unawareness score was correlated with all of the cognitive performance assessed on the five DRS subscales, with correlations from −0.34 to −0.55. The general unawareness score is most strongly correlated with the observed initiation score on the DRS (r = −0.55). This demonstrates the relevance of the MISAwareness method for correct representation of the relationship between unawareness and executive functioning.

The unawareness scores correlated with the global cognitive performance assessed on the DRS (r < −42; P < 0.001), except for unawareness of memory deficit (r = 0.20; P = 0.11).

More specifically, correlations between the five cognitive subscales and the unawareness of deficits for attention, initiation, construction, and conceptualization all fell between −0.18 and −0.49. The greater the degree of unawareness, the lower the participant's cognitive performance on the DRS. Correlations with unawareness of memory deficit, however, were lower, between 0.02 and −0.26, except for initiation score (r = −0.26; P < 0.05).


From a cognitive perspective, the present study shows the value of a method using prediction performance discrepancies to better assess anosognosia. The pattern of findings shows significant between- and within-group differences between predicted and observed scores. Comparison participants gave higher self-ratings than their performance for the overall DRS score. These results suggest that people are likely to overestimate their performance on cognitive tasks. They display biases in self-relevant assessments relating to general cognitive efficiency in such a way that a weak degree of unawareness seems to be a normal bias of the self-appraisal process. AD patients rated their cognitive functioning more highly than their performance for all subscales, minimizing the severity of dementia. Patients' self-rated predicted scores would be as high as comparison participants' scores if they were defending through denial. Unawareness in AD appears much more a difficulty of appreciating the severity of deficits in neuropsychological functioning than an impaired ability to recognize the presence of deficits. This result is also in agreement with those of the studies relying on a comparison paradigm between prediction and performance.[16, 19, 20] In the AD patients, all the differences between the predicted and observed performances are similar, as was the case for all the cognitive domains assessed, including memory. Conversely, only the memory awareness score is not correlated significantly with the global cognitive performance, while all of these correlations are notable for all the other awareness domains studied. This finding suggests a distinction between the mechanisms underlying the awareness of memory disorders and those underlying the other unawareness domains taken into account. It should also be noted that the only score in awareness of disorders significantly correlated with the awareness of memory disorders concerns the executive domain. This complexifies the data by suggesting a communality between the awareness mechanisms of the memory and executive disorders, a communality between the awareness mechanisms of the executive disorders and those underlying the other cognitive functions, but an independence between awareness of the memory disorders and those of the other cognitive functions taken into account. Among all the cognitive domains, only the executive performance score is correlated significantly with the score in awareness of memory disorders. The greater the degree of unawareness of memory disorders, the lower the participant's executive performance on the DRS. In most sectors investigated, none of the cognitive components suffices to explain disorder awareness. Conversely, the sole cognitive component that might partially participate in the alteration of awareness of memory disorders remains the executive component. A few explanatory elements of these disconnections may be provided by cognitive neuropsychology models. There are anosognosia theories applicable to AD originating from works on brain injury and which at the same time relate to injured sites and clinical phenomena observed. The common feature of these theories is that they rest on the postulate of a disruption of mechanisms that monitor the output of perceptual and cognitive modules.[21-24] The heterogeneity of the clinical manifestations of anosognosia is at the base of these theoretical explanations. Most of them put forward various forms according to the mechanism(s) involved at a central or specific level. The role of an executive dysfunction is compatible with (i) the correlations between awareness of the executive disorders and awareness of the other cognitive disorders; and (ii) the correlations between the executive performances and awareness of all the disorders, including, to a lesser extent, awareness of the memory disorders. The present findings simultaneously point to global awareness alteration (all the performances are overrated) but also an absence of a statistical link between awareness of the memory disorders and awareness of the disorders pertaining to the other cognitive functions, which conversely co-vary relative to one another, forming a coherent whole. If the executive dysfunction is common to both these forms of anosognosia, other phenomena also seem to be involved. Within anosognosia research, the object of awareness assessment may be restricted to a single domain, such as memory. In the present study, all unawareness scores were moderately inter-correlated except for memory. All unawareness scores were negatively correlated with cognitive performance except for unawareness of memory deficit. This result identifies a methodological difficulty, a conceptual limitation and raises the issue of the generalizability of studies solely focused on memory unawareness.

It is henceforth important to repeat this study with a large number of subjects so as to highlight more strongly the observed disconnections and possibly to identify others. The study of the disconnections can be extended in several ways, just like the global organization of the phenomena thus observed. Up to now, research has neglected the covariations between the different manifestations of anosognosia, not by denying the existence of these different phenomena but by designing the research as if the processes were one and the same. Jarring with this tendency, Starkstein's team has developed a standardized tool that identifies four distinct components of disorder awareness.[25] In the same manner, multidimensional analyses could be used to study the organization of the latent variables underlying the awareness of disorders as measured by MISAwareness.

Another tendency of research, deriving partially from the previous one, consists in considering all the patients as homogeneous concerning the awareness of disorders. This point nevertheless remains to be studied and it would be relevant to check for the existence of patients clusters, that is to say of awareness profiles of the disorders, which is today rendered possible both by significant numbers of subjects and by the existence of methods (e.g. Anosognosia Questionnaire for Dementia,[25] MISAwareness) identifying different domains of anosognosia. Misawareness must be compared with other paradigms using several different standards such as subjective judgment of a relative or clinical assessment of a professional caregiver. The measurements based on questionnaires may be more pregnant with self-efficiency beliefs while the measurements of prediction would be more directly relating to self-monitoring.[17] If disruptions between different cognitive dimensions can be observed, such disconnections may probably be present between the cognitive and affective or behavioral variables, just like prior findings.[1, 26, 27] The identification of all these disruptions and the putting in perspective of the different psychological processes of disorders awareness proper to each paradigm should enable the progressive outlining of the determinatives and the mechanisms of disorders awareness in AD. The model suggested by Clare provides the conceptual framework of such a scientific undertaking.[28] The identification of the different domains of disorders awareness, the consideration of different paradigms as reflecting the many underlying processes of the clinical manifestations of anosognosia and finally, on the basis of the current psychometric developments, the possible identification of patient clusters will allow for fine-tuning of the models, in a more integrative manner. It will then be possible to clarify the role of an executive dysfunction, whereas at the moment, it is not always easy to distinguish the relative contributions of selective frontal impairment compared with global cognitive deterioration to anosognosia.[18] Along these lines, the recent neuroimaging findings will support a complex picture and support previous neuropsychological data.[29]

Previous research found only weak correlations between a prediction-based and a questionnaire-based method.[30] This implies that these different methods represent different aspects of unawareness,[31] making a multidimensional model of unawareness a necessity. MISAwareness Assessment reflects a cognitive formulation of deficit unawareness. It is difficult to tease apart neurologically based symptoms from adaptive and social responses such as denial, belief or coping. There is a need to integrate this approach in a more comprehensive and psychosocial model of unawareness.


This work was supported by the Association France Alzheimer. The authors wish to thank Mélanie Lemaire and Florie Barois for their participation in the collection of data. This work has been developed and supported through the LABEX (excellence laboratory, program investment for the future) DISTALZ (Development of Innovative Strategies for a Transdisciplinary approach to Alzheimer disease).

No conflict of interest.