International Journal of Psychophysiology 38 Ž2000. 55᎐70 The application of cognitive event-related brain potentials ž ERPs/ in language-impaired individuals: review and case studies John F. Connolly a,b,U , Ryan C.N. D’Arcy a , Randy Lynn Newmana , Rachel Kemps c a Cogniti¨ er Clinical Neuroscience Unit, Department of Psychology, Life Science Centre, Dalhousie Uni¨ ersity, Halifax, NS, Canada B3H 4J1 b Department of Psychiatry, Dalhousie Uni¨ ersity, Halifax, NS, Canada B3H 4J1 c Catholic Uni¨ ersity of Nijmegen, Nijmegen, The Netherlands Received 1 September 1999; received in revised form 2 April 2000; accepted 2 May 2000 Abstract There is a substantial body of basic research that has utilized ERPs to investigate the neurological basis of cognition. This research has, in turn, led to the development of practical applications of cognitive ERPs in patient populations. In particular, recent work has focused on the development of ERP-based assessment measures for the neuropsychological assessment of dyslexia and language impairments secondary to stroke. This review describes the innovative assessment methods program ŽIAMP., an initiative to utilize ERPs for a neuropsychological assessment of patients who cannot be evaluated by traditional methods. The success of this program has demonstrated that ERPs can be used to reliably evaluate an individual’s reading and speech comprehension abilities, independent of behavioral and speech production impediments. In contrast to traditional neuropsychological assessment, these ERP methods can discern the cognitive strategies used by an individual to perform a task. 䊚 2000 Elsevier Science B.V. All rights reserved. Keywords: Event-related brain potentials; Neuropsychology; Assessment; Functional magnetic resonance imaging; Magnetoencephalography; Dyslexia; Stroke; Cognition; Language U Corresponding author. Tel.: q902-494-2961; fax: q902-494-6585. E-mail address: john.connolly@dal.ca ŽJ.F. Connolly.. 0167-8760r00r$ - see front matter 䊚 2000 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 7 - 8 7 6 0 Ž 0 0 . 0 0 1 3 0 - 6 56 J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 1. Introduction Patient assessment is the cornerstone of the caregiver’s choice of treatment, the process of therapeutic intervention, treatment monitoring and thus, ultimately, patient outcome. Evoked potentials ŽEPs. in response to simple auditory, visual or somatosensory stimuli have long been used to evaluate sensory function in patients ŽChiappa, 1997.. The success of this approach has led to the standard use of EPs in a variety of clinical contexts. The perception of EPs is that they reflect hardwired systems, and thus, are themselves reliable and invariant. It has long been held that while sensory EPs are reliable enough to be used in clinical contexts, cognitive event-related potentials ŽERP. may be too variable to be clinically useful. Another criticism has been that differences in ERPs between cognitive tasks can best be demonstrated with grand averaged waveforms and are less reliably demonstrated at an individual subject level. Simply put, this is a dated and fundamentally incorrect belief. ERPs provide an essential neurophysiological index of mental functions generally, and can be utilized to assess those functions following neurological insult on an individual subject basis. Connolly and D’Arcy Ž2000. have reviewed the development of and normative data for a series of ERP-based assessment measures of language function. This work has demonstrated that various neuropsychological tests can be adapted for computer presentations and simultaneous ERP recording. The presence or absence of the cognitive functionŽs. in question can be determined by the existence of a predicted ERP component, which has been shown previously to reflect that function. Why is the linking of ERP components to neuropsychological tests important? The ability to establish levels of intellectual function in patients who have experienced neurotrauma is essential for their assessment and treatment. However, a common consequence of neurological insult is impaired communication abilities ŽMarquardsen, 1969; Wade et al., 1986; Enderby and Philipp, 1987; Enderby et al., 1987; Morse and Montgomery, 1992; Pedersen et al., 1995.. Impaired communication frequently hinders the clinician’s ability to assess intellectual function, making neuropsychological testing difficult, unreliable or totally impossible. Thus, the availability of an assessment measure that functions independently of communication disabilities would provide invaluable information about intellectual function when traditional measures fail. It is possible to use ERPs in order to fulfill that role. Furthermore, when acquired in a psychometrically defined context Ži.e. using a neuropsychological test., the level of inference required to interpret a particular ERP waveform is minimal. For instance, in a language comprehension context, the failure to emit a P300 to a correct definition of a target word in a vocabulary test ŽConnolly et al., 2000,b. or a PMNrN400 to a spoken word incorrectly describing an accompanying picture in a picture᎐vocabulary test ŽByrne et al., 1995a,b; Connolly et al., 1995a. reflects a failure to comprehend the meaning of the target words. Confirmation of this interpretation has been demonstrated when vocabulary levels tested exceed the individual’s knowledge level and ERP components no longer differentiate correct from incorrect definitions. Aside from the applications that ERP-based measures have for assessing individuals with language impairments, these methods also have widespread implications for assessing attention, memory, perception, and levels of consciousness. In fact, any cognitive function currently measured with neuropsychological tests may be amenable to assessment with ERPs. The primary advantage of using ERP measures in addition to traditional neuropsychological measures is that the former provides information about the cognitive strategies used in performing a given task. This paper presents a review of this work and is divided into two sections. In the first section ŽSection 2., a subset of the neuropsychological tests that were adapted for computer presentation and simultaneous recording of ERPs and the ERP components associated with normal performance of those tests have been briefly reviewed. Please refer to Table 1 for a summary of these studies, and for a more detailed description of test adaptation and normative work, the interested reader is referred to Connolly and D’Arcy J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 57 Table 1 The table provides a summary of normative data from the adapted tests, an overview of the relevant cognitive processes, and a list of the associated ERP components Žsee text for further details. Paradigm Cognition ComponentŽs. References Semantic processing N400 Connolly et al., 1995b Phonological and semantic processing PMN, N400 Connolly and Phillips, 1994 Computerized PPVT-R Adults N s 12 Age: M s 27 Range s 23᎐34 Youth N s 3 Age: M s 17.6 Range s 16᎐18 Children N s 15. Age: M s 10.3 S.D.s 0.6 Receptive vocabulary level PMN, N400 P300 Byrne et al., 1995a,b Connolly et al., 1995a Computerized PALPA Adults N s 23 Age: M s 22.9 S.D.s 4.2 Reading comprehension P300 D’Arcy et al., 2000b Computerized Token Test Adults N s 27 Age: M s 23.2 S.D.s 5.8 Speech comprehension; attention and phonology P300, N2b D’Arcy and Connolly, 1999 D’Arcy et al., 2000a Psycholinguistic N400 Visual N s 20 Age M s 21 Range s 19᎐24 Auditory N s 20 Age: M s 21 Range s 18᎐43 Ž2000.. The purpose of the second section of this article ŽSection 3. was to highlight some of the recent applications of these tests in the evaluation of impaired language functions in dyslexic readers and stroke patients. 2. Normative ERP results in computerized neuropsychological tests 2.1. Peabody picture ¨ ocabulary test ᎏ re¨ ised (PPVT-R) The PPVT-R is a well-established assessment tool for evaluating vocabulary knowledge ŽDunn and Dunn, 1981.. The computer adaptation of the PPVT-R has maintained all critical aspects of the original test by dividing it into three levels of difficulty. For each item, participants viewed a picture paired with a spoken word Žthat was congruentrincongruent ., and the ERP recordings were time-locked to the onset of the spoken word. Fig. 1a represents the sequence and time course of stimulus presentation. The development of this version of the test was directed at establishing receptive vocabulary ERP norms in adults and children ŽByrne et al., 1995a,b; Connolly et al., 1995a.. Also, these initial studies compared per- 58 J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 Fig. 1. Ža. Schematic of the computerized PPVT-R representing the time course of stimulus presentation. Participants were required to judge whether or not a spoken word matched the picture. ERPs were time-locked to the verbal word onset Žarrow.. Žb. Grand average waveforms recorded at Fz, Cz, and Pz. The three levels corresponded to gradients of difficulty, such that ‘easy’ represented a pre-school vocabulary level, ‘moderate’ represented a child vocabulary level, and ‘difficult’ represented an adult vocabulary level. ERPs to incongruent picture᎐word pairs were differentiated from those to congruent picture-word pairs on the basis of the PMN and N400. Time in milliseconds Žms. is represented on the horizontal axis and amplitude in microvolts Ž␮V. on the vertical axis. Positive is up. ŽFigure adapted from Connolly et al., 1995a. J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 formance on the computerized version with that on the standard version, to establish that the adapted version had not violated the test’s psychometric properties. As Fig. 1b depicts, two distinct components were generated to the incongruent spoken words. First, the phonological mismatch negativity ŽPMN. was elicited when the initial phoneme mismatched that of the expected initial phoneme of the spoken word. Secondly, the N400 was elicited to violations in semantic expectancy. The PMN likely reflects sub-lexical spoken word processing, while the N400 reflects post-lexical processing in both speech processing and reading ŽConnolly and Phillips, 1994; Connolly et al., 1995b; Brink et al., 1999; Hagoort and Brown, 1999.. Thus, it is possible to measure not only comprehension, but also the stages of linguistic processing in patients on the basis of ERPs Žand independent of behavior.. More importantly, these effects were seen only if the word was within the individual’s vocabulary knowledge level. If the participant was tested at vocabulary levels beyond their knowledge, then the waveforms failed to differentiate. Also, differential patterns could be established at an individual subject level; an essential component of the process if the measures are to be used clinically. The PPVT-R adaptation represented one of the first applications of the IAMP in a clinical setting. A single case of an individual with cerebral palsy was used to evaluate whether ERPs could adequately track receptive vocabulary levels in such a patient. Briefly, data were recorded from a patient with cerebral palsy ŽCP. and three matched control subjects using the computeradapted PPVT-R ŽByrne et al., 1995a. As was anticipated, the healthy controls showed clear N400 differentiation at lower vocabulary levels, but did not do so at the higher level vocabulary levels because the words were beyond their knowledge levels and they were unable to judge when a word was incongruent to its accompanying picture. These findings were confirmed with their performance on the alternate version of the standard test. Significantly, the CP patient demonstrated precisely the same patterns as the matched 59 controls on both the ERP computerized test and on the standard test. 2.2. Psycholinguistic assessments of language processing in aphasia (PALPA) The written sentence comprehension section ŽTest 56. of the PALPA represents a test of sentence-reading comprehension that was designed from a cognitive neuropsychological, theory-driven orientation, to not only serve as a clinical tool, but also as a research device ŽKay et al., 1992, 1996.. The test assesses reading comprehension using a picture᎐sentence matching task. The format of the test uses a set of three pictures accompanied by a sentence that describes only one of them. The entire 60 picture᎐sentence sets in the standard test were adapted for computer presentation and ERP recordings ŽD’Arcy et al., 2000b.. The computerized version of the test used a study phasertest phase format. In the study phase, participants saw a set of three pictures and the accompanying sentence with the instruction to read the sentence and decide which picture it correctly described. The pictures and sentence went off the screen and were followed by the presentation of each picture separately. ERPs were recorded for each picture presentation. As Fig. 2a illustrates, correct pictures were differentiated from incorrect pictures on the basis of a P300 component. The P300 is an extensively investigated positive-going ERP component that has a variable latency depending on the complexity of the eliciting task and the processing speed of the individual. The P300 is conceptualized as reflecting a memory updating process in response to stimulus-driven changes in memory representations ŽPolich and Kok, 1995.. Picton Ž1992. has observed that, amongst other things, the P300 was able to determine whether an individual was processing stimuli. The parietally distributed P300 following correct, but not incorrect pictures, was interpreted to reflect a successful performance on the PALPA. Again, the critical issue was the test’s utility at the individual participant level. The P300 differentiated correct from incorrect pictures at the individual participant level with 91.3% accuracy. 60 J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 Fig. 2. Ža. ERPs obtained from a representative participant while selecting the picture that corresponded with a written sentence ŽPz.. There was a significantly larger P300 response Žasterisk . to correct pictures than to incorrect pictures. Žb. Individual ERPs recorded from LG, a dyslexic participant, in which no differentiation between conditions was apparent ŽPz.. All other details as in Fig. 1. ŽFigure adapted from D’Arcy et al., 2000b. In addition, participants’ Ž N s 23. performances on the standard PALPA and the computerized version were comparable with accuracy rates of 97.4 and 93.5%, respectively. These findings compare favorably with the norms provided by the test authors of 94.7%. The differential response pattern observed in the PALPA could be isolated further in reducedaverage Ž5-trial. waveforms within an individual. Data from a subset of participants Ž N s 10. were examined using event-resolution imaging ŽERI; Newman et al., 1999; Newman et al., in prepara- J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 tion.. Briefly, this technique formed response templates based on time, frequency, and spatial distribution for correct and incorrect trials. The resulting templates could then be applied to new data sets formed with low numbers of trials. The mean total accuracy, sensitivity, and specificity levels for the five trial average classification processes were 72.9% Ž12.2., 77.9% Ž12.0., and 74.6% Ž15.9., respectively. Thus, when a trial contained P300 activity Ži.e. a correct trial., ERI accurately detected and classified this trial as correct, 77.9% of the time. Equally, when an epoch did not contain P300 activity Ži.e. an incorrect trial., the trial was accurately classified as being incorrect 74.6% of the time. The ability of ERI to discriminate smaller sets of trials than what is currently capable with traditional averaging, offers the potential to more accurately denote trial types and gradients of difficulty within computerized neuropsychological tests. 2.3. Token Test The Token Test is used widely in neuropsychological contexts to evaluate speech comprehension in individuals ŽBoller and Vignolo, 1966; De Renzi and Vignolo, 1962; Lezak, 1995.. The test’s status as a gold standard for assessing comprehension made it ideal for computer adaptation and simultaneous ERP recording ŽD’Arcy and Connolly, 1999.. In the traditionally administered test, an array of geometric shapes is placed in front of the patient and hershe is instructed to, for example, ‘touch the green circle’ or ‘touch the black circle with the red square.’ For the purposes of the ERP testing, the usual test procedure was reversed. The computerized Token Test ŽCTT. presented an array of colored geometric shapes on a computer screen and an animated pointer moved amongst the shapes touching either one or two shapes. A digitized spoken sentence, describing the previous animation incorrectly or correctly Ž0.5 probability. followed this sequence. Participants were instructed to watch the animation, listen to the spoken description and indicate behaviorally whether the description was correct or incorrect. The incorrect spoken description contained one word that did not match the ani- 61 mation. ERPs were recorded at the onset of this word. Thus, for example, if the pointer had touched a black square, the description may have said either ‘black circle’ Žincorrect. or ‘black square’ Žcorrect.. The mismatching words could occur in either the shape word or the color word. Also, as usual, a version of the Token Test Žthe ‘short version’, De Renzi and Faglioni, 1978. was administered to all participants in order to evaluate their performance on the computerized version with the standard version administered traditionally. Performance on the standard and computerized test proved to be comparable, with percentage correct scores of 97.1 and 96.8%, respectively. These values compared favorably with other data for normal adult populations of 98.7% ŽSpreen and Strauss, 1991.. ERPs were found to differentiate incorrect and correct spoken descriptions of the animations seen on the CTT ŽD’Arcy and Connolly, 1999.. Centro-parietal P300 responses were seen at approximately 500 ms after the presentation of the incorrect verbal description. The delayed latency reflected the complexity of the processing required to complete the task. Such cognitive processing is, by nature, variable ŽGevins and Cutillo, 1986. and reflects a variety of functions, not least of which is stimulus evaluation time ŽPritchard, 1981.. Also, the processing of a speech signal takes place over a period of time, contributing further to the delay in the P300 latency. The P300 differentiation between incorrect and correct descriptions was seen reliably in over 92% of the participants, confirming the utility of this procedure at the single case level. Participants performed the test in a ‘target detection’ manner that required speech comprehension abilities and attention ŽD’Arcy and Connolly, 1999; D’Arcy et al., 1999, 2000a.. Furthermore, it has recently been reported that the N2b1 can be elicited following a deviation in an elaborate speech processing task, demonstrating that this component is also sensitive to violations 1 The N2b is associated with voluntary attentional mechanisms used to detect stimulus deviance in processing sensory and cognitive stimuli ŽD’Arcy et al., 2000a.. 62 J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 in an active template formed from sensory processes ŽRitter et al., 1992; Naatanen, 1995. ¨¨ ¨ and cognitive processes ŽD’Arcy et al., 2000a.. Additionally, the authors speculated that the cognitive template was formulated on the basis of phonological working memory ŽBaddeley, 1992., which maintained and analyzed the visuospatial information from the previous animation. With respect to clinical applications, these findings demonstrated that in many of the computerized adaptations of neuropsychological tests, more than one ERP component differentiates correct from incorrect trials Žsee also, Connolly et al., 1999a.. Furthermore, and of particular importance, these findings demonstrated that, by utilizing well-known relationships between ERP components and cognition, the manner in which an individual performs a neuropsychological task can be established. 2.4. Psycholinguistic N400 paradigm This experimental paradigm has been used in a variety of studies over the last 5 years Že.g. Connolly et al., 1995b; Helenius et al., 1998, 1999.. The initial experiment using this design determined that the PMN and N400 were, in fact, distinct waveforms representing distinct language-related functions ŽConnolly and Phillips, 1994.. The design used four conditions that varied along both phonological and semantic dimensions by manipulating the terminal words of contextually-constrained sentences. One condition involved the use of a sentence-terminal word that was the expected, semantically correct ending: ‘The piano was out of tune.’ The resulting ERP showed only a large, slowly developing positivity with no indication of either the PMN or N400. A second condition used sentences ending with a word that violated semantic expectancy and has an initial phoneme that is also unexpected: ‘Joan fed her baby some warm nose’ Žmilk being the expected word. produces both a PMN and a N400 in response to the terminal word. The PMN was seen in response to the initial phoneme mismatching the expected initial phoneme of the word milk, while the N400 reflected the semantic inappropriateness of the word. The remaining two conditions are those that were designed to produce a double dissociation between the PMN and N400. Thus, ‘The gambler had a bad streak of luggage’ Žluck being the expected word. resulted in a N400 with no preceding PMN because the word presented Žluggage. shared its initial phoneme with the expected word Žluck. but was semantically inappropriate to the sentence context. The final sentence type ‘The pigs wallowed in the pen’ Žmud being the expected word. had an initial phoneme that mismatched context-driven expectations but was, in fact, semantically acceptable. Thus, the terminal words in these sentences produced a PMN but no subsequent N400. The preceding section has provided an overview of a selection of paradigms that can be used to evaluate cognitive functions. In the Section 3,these paradigms are utilized in a series of individuals with language impairments Ždyslexia and stroke. selected to demonstrate the efficacy of the IAMP and profile its use of ERPs as neurophysiological indices of cognition. 3. ERP results for clinical populations with language impairments 3.1. Dyslexia Recent work has utilized various neuroimaging techniques to examine language processing in dyslexic readers. Helenius et al. Ž1999. used the psycholinguistic N400 paradigm ŽConnolly et al., 1995b. to record event-related magnetic fields ŽERMF. and ERPs in normal and dyslexic readers and to examine differences in the spatio-temporal characteristics of the N400m2 . While a N400m was observed for both groups with no significant difference in the spatial distributions, the dyslexic readers showed differences in the timing and the strength of activation. In particular, the N400 response was delayed by ; 100 ms and the amplitude was reduced Žthis reduction was especially noticeable following orthographic manipulations.. Accordingly, the authors hy- 2 The N400m is the magnetic equivalent of the N400. J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 pothesized that the dyslexic subjects were more reliant on sublexical word recognition Ži.e. orthographic analysis . and, thus, may have utilized a smaller or less synchronous neural population during whole word semantic analysis. While the localization of N400m sources to the left superior temporal gyrus was similar for both groups, the pattern was much less distributed for the dyslexics. These results were relevant to data obtained in two independent case studies, described below, which outlined ERP manifestations of atypical language function and related fMRI findings. 3.2. Case study: LG In research that used ERPs for the assessment of reading ŽPALPA. and speech comprehension ŽCTT., the examination of neural responses in the case study, LG, provided preliminary evidence for the sensitivity of ERPs to subtle developmental difficulties with reading Žunpublished data.. LG, a third year university undergraduate, was diagnosed with a learning disability at the age of eight and reported ongoing difficulties with reading, but not speech comprehension. Prior to university, he had received special training for his reading difficulties. In light of these reading-specific problems, it was predicted that his ERPs on the PALPA would be abnormal, but the expected differentiation Ži.e. normal. would occur on the CTT. Recall that the ERPs for healthy controls reliably differentiated correct and incorrect test items on the basis of the P300 response for both the PALPA and the CTT paradigms. In contrast, Fig. 2b illustrates that LG showed no observable differentiation between the correct and incorrect test items on the PALPA. Further support that the waveforms failed to differentiate conditions was found using a finer-grained signal analysis with ERI ŽNewman et al., 1999; Newman et al., in preparation.. The analysis determined that 100% of correct trials were classified as being incorrect trials Ži.e. no differentiating signal could be modeled.. However, it was clear that he was processing the stimuli because the N1᎐P2᎐N2 complex was clearly identifiable and, of particular importance, his behavioral scores were perfect. Inter- 63 estingly, his ERPs on the CTT differentiated the incorrect and correct test items and the response pattern was comparable to that of other subjects. Thus, only the ERP findings corroborated LG’s initial report of difficulties with reading, but not speech comprehension. Two important observations resulted from LG’s performance and ERP findings. First, his behavioral scores were high, and yet on the PALPA the ERPs did not concur with the behavioral results. This discrepancy did not indicate that the ERPs were unreliable Žas the measure has been proven in matched controls and the waveforms indicated intact stimulus processing., but instead suggested that LG performed this reading task Ži.e. identified the correct test items. in a different manner than normal readers. D’Arcy et al. Ž2000b. have stated that the normal readers were capable of formulating accurate linguistic representations and then subsequently identifying the correct test item that matched this linguistic representation through recognition memory processes. This process was inferred through the examination of the task demands and the ERP waveform, specifically the presence of the P300. It is likely that the observed absence of the P300 response for LG indicated a significantly different processing strategy. For instance, if LG experienced difficulty in either the formulation or maintenance of linguistic representations, then the impaired linguistic representation may have required a more intensive item-by-item decision process instead of a recognition process directed at identifying a previously selected picture. Thus, this item-by-item decision process would be directed at discriminating both the correct and incorrect test items during the data collection phase. In support of this interpretation, LG showed a strong negative-going waveform Ž200᎐500 ms period. in both experimental conditions. This negative-going waveform was consistent with the ERP patterns hypothesized to reflect active linguistic processing during the computerized PPVT-R ŽByrne et al., 1995a,b; Connolly et al., 1995a.. The second important observation concerned the sensitivity of ERP responses during the assessment of cognitive functions. In the case of LG, the results from the neuropsychological test 64 J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 did not reflect any difficulties with reading, and using these results alone it would not have been possible to identify the anomalies associated with this case. It was only when ERPs were simultaneously recorded, that it was possible to detect differences secondary to a reading disability. However, an important caveat exists; these findings were based on a single case study and further evidence for these ERP assessment techniques must be obtained from other high-functioning dyslexic individuals. 3.3. Case study: BP Recently, ERPs and fMRI have been used to evaluate reading abilities in a high-functioning individual ŽBP. with profound dyslexia. Traditional neuropsychological testing revealed that BP had a severe phonological impairment as measured by non-word reading and phoneme deletion tasks. Despite this reading-related impairment, BP had developed successful compensatory strategies as demonstrated by his ability to obtain a Master’s degree. Using ERPs, BP was tested on the psycholinguistic N400 paradigm ŽConnolly et al., 1995b., the computerized PALPA, and the CTT. Fig. 3 depicts BP’s performance on each of these tests as compared to a representative control participant. On the psycholinguistic N400 paradigm, BP showed no differentiation whatsoever following the presentation of anomalous terminal words in which the orthographicrphonological andror semantic features were unexpected. The failure to observe a N400 suggested that BP had a reduced capacity to process the lexicalrsub-lexical features of the terminal word and had difficulty integrating it into the context of the preceding sentence. In support of this interpretation, BP also showed no ERP differentiation to the PALPA test items. ERI analysis found that 83% of incorrect trials were classified as correct trials, indicating that there was insufficient signal differentiation between the two conditions. Indeed, the findings in this case replicated the previous case study in that BP’s behavioral performance was accurate. In addition, when BP was tested on speech comprehension with the CTT, there was an observable differentiation in the waveforms Žagain, similar to LG’s results .. The only difference between the two case studies involved the morphology of the waveforms. In the case of BP, there appeared to be P300 responses in both conditions, whereas, in LG’s data there was a sustained negativity. Nonetheless, in both cases, it was likely that the lack of expected differentiation resulted from an item-by-item decision strategy Žmaking all test items equally salient .. As mentioned previously, both ERP and fMRI data were collected for BP. The fMRI Ž1.5 T. provided a neuroanatomical backdrop for the ERP components used in language assessment with the psycholinguistic N400 paradigm. Fig. 4a highlights the contrasting pattern of activation observed between a representative subject and BP. In normal readers, two key regions were activated during reading in general: Ž1. the visual cortex Žstriate and extra striate .; and Ž2. Broca’s area. The left perisylvian region Žangular gyrus., left posterior parietal regions, posterior cingulate gyri, and frontal areas were particularly sensitive to the psycholinguistic properties of the terminal word ŽD’Arcy et al., 1999; Connolly et al., in preparation.. In contrast, the reading-related profile obtained from BP differed markedly. While BP did show activation in the visual cortex, the remaining regions identified in controls were not activated. Instead, robust activation was observed centered in the sylvian fissure Žanterior to the primary auditory cortex. which, while bilateral, was stronger in the left hemisphere. Fig. 4b demonstrates the overlap between active regions identified with fMRI and MEG N400m dipole solutions ŽHelenius et al., 1999. for high functioning dyslexics. Considering both the ERP and fMRI results, it was likely that BP utilized a different reading strategy, which in turn was evidenced by differences in neural processing and the anatomical regions of activation. 3.4. Stroke As the objective of the IAMP was to use ERPbased measures to assess non-communicative J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 65 Fig. 3. Individual ERPs at Pz obtained from BP, a dyslexic on the psycholinguistic N400 paradigm Ža., the PALPA Žb., and the CTT Žc.. Individual ERPs obtained from representative control participants on each of the respective tests are inserted for comparison. Ža. BP failed to show an observable N400 to anomalous sentence-ending words. Waveforms corresponding to each of the four conditions are represented as follows: semantically congruentrphonologically expected ŽRR., semantically congruentrphonologically unexpected ŽWR., semantically incongruentrphonologically expected ŽRW., semantically incongruentrphonologically unexpected ŽWW.. Žb. BP showed no differentiation on the PALPA ŽP300 in both conditions.. In contrast, a control participant showed a larger P300 in the correct condition. Žc. On the CTT, ERPs differentiated incorrect from correct spoken descriptions of the animation sequence for both the control participant and BP. The P300 is denoted by an asterisk. All other details as in Fig. 1. 66 J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 Fig. 4. Ža. Functional MRI Žblood oxygen level dependent contrasts at 1.5 T. depicts the regions activated during the psycholinguistic N400 paradigm. For a normal reader, robust reading-related Žgray checker board. activation was observed in the visual cortices Žstriate and extrastriate . and Broca’s area. In addition, activation related to evaluating the terminal word Žcross hatched. was observed in the left angular gyrus, posterior cingulate gyri, and frontal regions Žcircles.. In contrast, BP showed predominant activation in the visual cortex and the left sylvian fissure region Žcircle.. Note that fMRI images are presented radiologically with an AC᎐PC orientation for the control and a horizontal orientation for BP Ži.e. visual cortex is not imaged for BP.. Žb. Source localization of N400m dipoles in the left and right hemispheres for controls and dyslexics, from the same paradigm. Both the controls and dyslexics showed activation in the left superior temporal gyrus, though sources for the dyslexics were more focused. ŽFigure adapted from Helenius et al., 1999.. J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 patient populations, a group of stroke patients became the focus of recent work. The patients described below were assessed using the Boston Diagnostic Aphasia Examination ŽGoodglass and Kaplan, 1983.. Patient 1 was a 53-year-old woman who suffered a left hemisphere stroke 1 month prior to her participation in the study. She showed mild expressive aphasia and apraxic speech, the combination of which presented as stuttering. Her reading comprehension was unimpaired at the sentence level, but her reading became laborious as the complexity of material increased. Patient 2 was a 71-year-old man who suffered a left lacunar stroke 1 month prior to testing. He showed mild impairments on measures of oral expression, reading comprehension, and fluency, and severe impairments on measures of written expression. He also presented with mild dysarthria, mild word finding difficulties and difficulty attending to lengthy text during reading. Patient 3 was a 60-year-old female who suffered a left hemispheric stroke 2 months prior to testing. Of the three patients tested, she exhibited the most language-based difficulties. She presented with moderate expressive, and mild to moderate receptive difficulties. She showed perseverations, paraphasias, neologisms, and decreased meaningfulness in the content of her output. While her auditory comprehension was unimpaired at the sentence level, her reading comprehension was moderately impaired at the sentence level. The three patients were tested on the standard and computerized versions of the PALPA. Data from these patients were compared to a group of elderly controls, which like young adults ŽD’Arcy et al., 2000b., showed ERP differentiation as indexed by a P300 to correct pictures. Importantly, the controls demonstrated waveform differentiation regardless of whether or not they provided a behavioral response. Scores obtained on the standard version of the PALPA were 83.3, 90, and 76.7% for patients 1, 2 and 3, respectively. The mean score for the elderly controls was 97%, indicating that all patients performed more than 1 S.D. below the mean of the control group. As shown in Fig. 5, the ERPs 67 of patients 1 and 2 on the computerized version showed a differentiation between correct and incorrect conditions, while patient 3 showed no significant differentiation. Furthermore, the amplitude of the P300 to correct pictures matched the pattern of patient performance on the standard PALPA, i.e. patient 2, who was most accurate on the behavioral PALPA, exhibited the largest P300. Patient 1, whose behavioral score was intermediate between patients 2 and 3, demonstrated reduced P300 activity in comparison to patient 2. Finally, patient 3, who had the lowest behavioral score, showed no evidence of P300 activity. Two important points about these results should be highlighted. Firstly, ERPs were able to index comprehension of the written text, irrespective of whether a behavioral response was required. The ability to record ERPs to stimuli that invoked cognitive processing regardless of the presencerabsence of behavioral responding enables this technique’s use with non-communicative and ‘difficult to assess’ individuals Žsee also Connolly et al., 1999b.. The second important point is that ERP differentiation has been shown to occur, not only in healthy, young undergraduate volunteers, but also in patients of widely differing ages who have a variety of disabilities. The bottom line is that this method works in real life situations. 4. Conclusion Cognitive ERPs are capable of monitoring a wide variety of cognitive processes in a reliable and valid manner with or without accompanying behavioral responding. They have the demonstrated the ability to serve clinical assessment, monitor therapeutic progress, and assist in prognosis. In fact, ERPs can provide information in situations where no other method can, such as with patients experiencing severe communication and motor impairments. They can also serve as adjunctive methods in situations that permit traditional assessment as they provide a unique level of information capable of informing the clinician of problems where traditional tests do not Žas seen with the LG case study above.. While neu- 68 J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 Fig. 5. Individual ERPs obtained on the PALPA from three stroke patients ŽPz.. Patients 1 and 2 both showed significantly larger P300 responses Žasterisk. to correct pictures than to incorrect pictures. In contrast, patient 3 failed to show observable differentiation between conditions. All other details as in Fig. 1. ropsychological tests may depend on a variety of cognitive processes, they are generally partitioned on the basis of the major functionŽs. they purport to measure. The behavioral measurements ob- J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 tained from these tests are limited in their ability to delineate these underlying cognitive processes. The use of ancillary ERP methods provide an on-line index of the cognitive processes involved in a particular task and are paramount to determining the individual strategies used. Techniques such as those encompassed in the Innovative Assessment Methods Program represent an essential development in clinical monitoring particularly when health care systems in many regions are under financial pressure. It is also a situation that will steadily worsen as baby boomers reach ages of increasing risk for stroke and Alzheimer’s disease, for example. Many therapeutic interventions Že.g. various rehabilitation procedures. involving neurological insult cases carry with them extended costs due to extended therapy periods and intensive personnel involvement. There is an understandable reluctance to commit resources to patients who may not be able to benefit from certain procedures. In most situations, patients who are incapable of communication and, therefore, who are incapable of being assessed, fall, by default, into that category of patients who are judged to show no signs of benefiting from therapeutic intervention. This is a situation experienced by a variety of patient groups from cerebral palsy ŽByrne et al., 1995a. to traumatic brain injury ŽConnolly et al., 1999b.. The latter case study is particularly germane and merits a brief description to emphasize the impact the techniques described above can have on a patient’s future. HF was a 21-year-old man who suffered a traumatic brain injury as a result of being stabbed through the head with a knife. Having survived the attack, he was admitted to a rehabilitation setting for assessment. The severity of his injuries rendered him globally aphasic and physically impaired to a level that prevented him from making reliable movements of any type. Traditional assessment could not be conducted and rehabilitation was not recommended as he appeared to have no apparent cognitive functioning abilities. He was referred for ERP testing and was found to have intact speech comprehension abilities despite being unresponsive to all verbal challenges. On the basis of these ERP findings, he was admit- 69 ted to a rehabilitation program that culminated in an excellent outcome. Without such intervention, this patient would have failed to return to a productive life, presuming he survived at all. Research over the last 35 years has demonstrated the sensitivity of ERPs to a plethora of cognitive functions and current research is extending these boundaries continuously. The time has arrived to apply this extensive knowledge base to those individuals experiencing cognitive problems due to various neurological insults including stroke, traumatic brain injury, and neurodegenerative disorders. The research encompassed in the Innovative Assessment Methods Program aims to do precisely that. Acknowledgements This research was supported by grants from the Scottish Rite Charitable Foundation of Canada, the Natural Sciences and Engineering Research Council of Canada, and the Heart and Stroke Foundation of Canada. Thanks are extended to A. Major, D. Cook, and N. Cook for assistance with the data reviewed in this article. References Baddeley, A., 1992. Working memory. Science 255, 556᎐559. Boller, F., Vignolo, L.A., 1966. Latent sensory aphasia in hemisphere-damaged patients: an experimental study with the Token Test. Brain 89, 815᎐831. Brink, D., van den Brown, C.M., Hagoort, P., 1999. The N200 component as an electrophysiological indicator of early contextual influences in spoken word recognition. Submitted for publication. Byrne, J.M., Dywan, C.A., Connolly, J.F., 1995a. An innovative method to assess the receptive vocabulary of children with cerebral palsy using event-related brain potentials. J. Clin. Exp. Neuropsychol. 17 Ž1., 9᎐19. Byrne, J.M., Dywan, C.A., Connolly, J.F., 1995b. Assessment of children’s receptive vocabulary using event-related brain potentials. Child Neuropsychol. 1 Ž3., 211᎐223. Chiappa, K.H., 1997. Evoked Potentials in Clinical Medicine, 3rd ed. Lippincoff-Raven, NewYork. Connolly, J.F., Phillips, N.A., 1994. Event-related potential components reflect phonological and semantic processing of the terminal word of spoken sentences. J. Cogn. Neurosci. 6, 256᎐266. 70 J.F. Connolly et al. r International Journal of Psychophysiology 38 (2000) 55᎐70 Connolly, J.F., Byrne, J.M., Dywan, C.A., 1995a. Assessing adults receptive vocabulary with event-related brain potentials: an investigation of cross-modal and cross-priming. J. Clin. Exp. Neuropsychol. 17 Ž4., 548᎐565. Connolly, J.F., Phillips, N.A., Forbes, K.A., 1995b. The effects of phonological and semantic features of sentence-ending words on visual event-related brain potentials. Electroencephalogr. Clin. Neurophysiol. 95, 276᎐287. Connolly, J.F., D’Arcy, R.C.N., 2000. Innovations in neuropsychological assessment using event-related brain potentials, Int. J. Psychophysiol. 37 Ž1., 31᎐47. Connolly, J.F., Major, A.M., Allen, S.L., D’Arcy, R.C.N., 1999a. Performance on WISC-III and WAIS-R NI vocabulary subtests assessed with event-related brain potentials: An innovative method of assessment. J. Clin. Exp. Neuropsychol. 21 Ž4., 444᎐464. Connolly, J.F., Mate-Cole, C.C., Joyce, B.M., 1999b. Global aphasia: an innovative assessment approach. Arch. Phys. Med. Rehabil. 80, 1309᎐1315. Connolly, J.F., D’Arcy, R.C.N., Gates, L., Service, E. An investigation of word and context comprehension using event-related functional magnetic resonance imaging ŽfMRI.. ŽIn preparation.. D’Arcy, R.C.N., Connolly, J.F., 1999. An event-related brain potential study of receptive speech comprehension using a modified Token Test. Neuropsychologia 37 Ž13., 1477᎐1489. D’Arcy, R.C.N., Connolly, J.F., Gates, L., Service, E., 1999. An investigation of word and context comprehension using event-related functional magnetic resonance imaging ŽfMRI.. Psychophysiology 36, s3. D’Arcy, R.C.N., Connolly, J.F., Crocker, S.F., 2000a. Latency shifts in the N2b component track phonological deviations in spoken words. Clin. Neurophysiol. 111 Ž1., 40᎐44. D’Arcy, R.C.N., Connolly, J.F., Eskes, G.A., 2000b Evaluation of reading comprehension with neuropsychological and event-related brain potential ŽERP. methods. J. Int. Neuropsychol. Soc. 6, 552᎐563. De Renzi, E., Faglioni, P., 1978. Normative data and screening power of a shortened version of the Token Test. Cortex 14, 41᎐49. De Renzi, E., Vignolo, L.A., 1962. The Token Test: a sensitive test to detect disturbances in aphasics. Brain 85, 665᎐678. Dunn, L.M., Dunn, L.M., 1981. Peabody Picture Vocabulary Test-Revised. American Guidance Service, Minnesota. Enderby, P., Philipp, R., 1987. Speech and language handicap: towards knowing the size of the problem. Brit. J. Disord. Commun. 21, 151᎐165. Enderby, P., Wood, V.A., Wade, D.T., Hewer, R.L., 1987. Aphasia after stroke: a detailed study of recovery in the first 3 months. Int. Rehabil. Med. 8, 162᎐165. Gevins, A.S., Cutillo, B.A., 1986. Handbook of Electroencephalography and Clinical Neurophysiology. Elsevier, Amsterdam. Goodglass, H., Kaplan, E., 1983. Boston Diagnostic Aphasia Examination ŽBDAE.. Lea and Febiger, Philadelphia. Hagoort, P., Brown, C.M., 1999. ERP signatures of listening to speech: I. Semantic effects. Žsubmitted for publication.. Helenius, P., Salmelin, R., Service, E., Connolly, J.F., 1998. Distinct time courses of word and context comprehension in the left temporal cortex. Brain 115, 1769᎐1782. Helenius, P., Salmelin, R., Service, E., Connolly, J.F., 1999. Semantic cortical activation in dyslexic readers. J. Cogn. Neurosci. 11 Ž5., 535᎐550. Kay, J., Lesser, R., Coltheart, M., 1992. Psycholinguistic Assessments of Language Processing in Aphasia. Lawrence Erlbaum Associates, East Sussex, UK. Kay, J., Lesser, R., Coltheart, M., 1996. Psycholinguistic assessments of language processing in aphasia ŽPALPA.: an introduction. Aphasiology 10 Ž2., 159᎐215. Lezak, M., 1995. Neuropsychological Assessment, 3rd ed. Oxford University Press, NewYork. Marquardsen, J., 1969. The natural history of acute cerebrovascular disease. A retrospective study of 769 patients. Acta. Neurol. Scand. Suppl. 38, 67᎐68. Morse, P.A., Montgomery, C.E., 1992. Neuropsychological evaluation of traumatic brain injury. In: White, R.F. ŽEd.., Clinical Syndromes in Adult Neuropsychology: The Practitioner’s Handbook. Elsevier, Amsterdam, pp. 86᎐176. Naatanen, R., 1995. The mismatch negativity: a powerful tool ¨¨ ¨ for cognitive neuroscience. Ear. Hear. 16 Ž1., 6᎐18. Newman, R.L., Connolly, J.F., D’Arcy, R.C.N., Cook, D., Cook, N., 1999. Event-resolution imaging: overcoming the limitations of signal averaging in ERP research. Psychophysiology 36, s85. Newman, R.L., Connolly, J.F., D’Arcy, R.C.N., Cook. D., Cook, N. Event-resolution imaging: an overview of methodology and implications for ERP research. ŽIn preparation.. Pedersen, P.M., Jørgensen, H.S., Nakayama, G., Raaschou, H.O., Olsen, T.S., 1995. Aphasia in acute stroke: incidence, determinants, and recovery. Ann. Neurol. 38, 659᎐666. Picton, T.W., 1992. The P300 wave of the human event-related brain potential. J. Clin. Neurophysiol. 9 Ž4., 456᎐479. Polich, J., Kok, A., 1995. Cognitive and biological determinants of P300: an integrative review. Biol. Psychol. 41, 103᎐146. Pritchard, W.S., 1981. Psychophysiology of the P300. Psychol. Bull. 89 Ž3., 506᎐540. Ritter, W., Paavilainen, P., Lavikainen, J. et al., 1992. Eventrelated potentials to repetition and change of auditory stimuli. Electroenceph. Clin. Neurophysiol. 83, 306᎐321. Spreen, O., Strauss, E., 1991. A Compendium of Neuropsychological Tests: Administration, Norms, and Commentary. Oxford University Press, New York. Wade, D.T., Hewer, R.L., David, R.M., Enderby, P.M., 1986. Aphasia after stroke: natural history and associated defects. J. Neurol. Neurosurg. Psychiat. 49, 11᎐16.