426 Electr~nceFhalograFh~and clinical Neurophysiologv, 1986; 63: 426-430 NEUROMETRICS AND DJ Elsevier Scientific Publishers Ireland. Ltd. control group of 57 normal re Short communication old at the time of this study. ing criteria designed 10 emu between the two groups, U NEUROMETRICS DOES NOT DETECT ‘PURE’ DYSLEXICS psychological conditions whit ings unrelated to dyslexia p C.D. YINGLING *, D. GALIN *, G. FEIN *.**, D. PELTZMAN and L. DAVENPORT * for our control group were * University of California, Sun Francisco, CA 94143, and ** San Francisco VA Medical Center. Sun Francisco, CA 94121 (V.S.A.) scribed by John et al. (198 more severely disabled than I below). Our sample is ti (Accepted for publication: November 26. 1985) (see assess the validity of the ne brain dysfunctions specificd ties. We were able to remnt Summary Thirty-eight severely dyslexic boys and 38 good readers were evaluated with neurometrics, a diagnostic prowdurr after their original testing, based on the application of numerical taxonomy to EEG spectra obtained during resting conditions, supplemented by selected e\

0013-4649/86/%03.50 6 1986 Elsevier Scientific Publishers Ireland, Ltd. vEUROMETRICS AND DYSLEXIA 421

mntrol group of 57 normal readers, all between 10 and 16 years Procedures dd at the time of this study. All subjects met rigorous screen- All data were acquired with a Model NM-1OlA data ing criteria designed to ensure no overlap in reading abilities acquisition unit made available to the project by Neurometrics, hetween the two groups, and no coexisting neurological or Inc., New York, who further agreed to analyze the records psychological conditions which might cause physiological find- blind. The data acquisition unit conntrolled delivery of stimuli ,figs unrelated to dyslexia per se. Academic screening criteria and digitized 20 channels of data, which were stored on floppy for our control group were roughly equivalent to those de- disks. The bandpass of the amplifiers was 1.9-29.0 c/sec at 3 gibed by John et al. (1980). while our dyslexic group was dB down with a 60 c/sm 45 dB notch filter. Standard EEG more severely disabled than those in the Ahn et al. (1980) study electrodes were attached at all 19 locations of the International below). Our sample is thus an ideal test set with which to 10-20 system referenced to linked ears, and the 20th channel assess the validity of the neurometrics procedure in detecting was used to record EOG. brain dysfunctions specifically associated with learning disabili- A standardized protocol developed by Neurometrics. Inc. lies. VI e were able to recontact most of the subjects, 1-3 years was followed throughout. The system was first calibrated for . a diagnostic procedurr after their original testing, and 40 controls and 38 dyslexics, artifact rejection based on each subject's characteristic EEG. ented by selected e\,,k& 69% of the original group, returned to participate in the present An artifact-free segment of eyes closed EEG was defined for ividual's values to tho% stud! each subject on the basis of predetermined criteria, such as normal controls (Ahn et Ail subjects were Caucasian, right-handed. middle class peak EOG values of less than SO pV and greater alpha power in boys An extensive history was taken from parents, including posterior than anterior leads. Subsequently, epochs in which ntellectual, neurological. pregnancy and perinatal complications, medical, social and the values exceeded those in the calibration segment by a ported earlier. Honcicr, academic problems. Children with histories of emotional prob- predetermined amount were automatically excluded. All experi- pups were classifled ar lems. hyperactivity, or birth stress were excluded. Neurological mental runs continued until 60 sec of artifact-free EEG (or 50 rted previously in niore screcning was carried out based on the examination methods of single trials for evoked potentials) had'been obtained; thus. the Touwen and Prechtl (1970) including cranial nerve functions, runs were of varying length depending on the amount of rnusde strength and symmetry, muscle stretch reflexes, superfi- artifact. Further artifact rejection was carried out by off-line cial reflexes, 5 categories of motor functioning (balance, tremor, programs. associated movements, abnormal movements, sequencing), and After artifact calibration, data were collected in a number perLeptual-motor performance including double simultaneous of standard conditions. always in the same order: eyes closed bados. The neurologicai tow h. two-point localization on fingers, graphesthesia, and EEG, 3 visual evoked potential conditions. and 2 auditory EP alities ('true positive.'), imilation of gestures. Exclusion criteria were moderate disabili- conditions. The 3 visual EP conditions were blank flash, coarse xitive rates of 46% and tie. in three or more-motor categories, or a consistent pattern checkerboard pattern onset, and fine checkerboard pattern n the original estirnare, 01 disability in the sensory-perceptual area. No child with onset (IS1 =I sec). The square screen subtended a viewing . combined with the I,>M millor neurologic signs, such as clearly impaired motor coor- angle of 6'. The auditory conditions were regular (l/sec) 45 dB that quantitative Et(; dination or seizures, was included. Pure-tone audiometry, visual SPL clicks and irregular clicks, in which the IS1 varied unpre- '1 both neurological d,,. acuity, refraction, accommodative amplitude, and binocular dictably with a mean IS1 of 1 sec, all delivered free-field from a Jbstantial proportion ,d function examinations were performed. All subjects had clini- small loudspeaker in front of the subject. Evoked potential ce of such findings &. cally normal hearing and vision. analyses were based on the monopolar montage as recorded; ialities can be shown 1,) All subjects had WISC-R full scale IQ above 88 with for EEG spectral analyses, a bipolar montage was generated by per se. or whether the\ di'ferences between verbal and performance scales less than 30 the computer to produce 8 channels: bilateral central (c3-cz. :urological deficits. N I' points. Groups were defined with a reading quotient: (2 X c4-C~).temporal (T3-TS. T4T6). parieto-occipital (P3-01, P4- e same procedures em- reading age)/(chronological age + mental age). Reading ages 02). and fronto-temporal (F7-T3. F8-T4). i&Y screened group nf uere based on the total score of the Gates-MacGinitie reading The resultant data set consisted of 1 min of eyes closed :d control group, all or test and on the Gray Oral reading test. Mental ages were based EEG, auditory EPs to regular and irregular clicks. and visual I dysfunction. We now on the WISC-R. All dyslexics had quotients less than 0.84 for EPs to blank flash and to coarse and fine checkerboard pat- '1 significant abnormall- both oral and Silent reading. Silent reading scores of controls terns. Data from 2 control subjects were lost due to a disk drive t no other clinical find- uere over 0.93, and their oral reading was within 1 year of problem; data disks from the remaining 38 dyslexics and 38 actual grade level, or higher. The two groups thus obtained, controls were encoded to reveal only the subjects' ages and although having equivalent ages and performance IQs, were were mailed to the Neurometrics, Inc. Data Analysis Center in n,idely divergent in reading ability. At the time of our original New York. which generated quantitative reports of EEG spec- screening, when the subjects were from 9 to 13 years old. tral features and certain EP indices and returned them to our control subjects averaged at the 8.5 grade level in both silent laboratory. and oral reading; dyslexics were at the 3.0 grade level in silent reading and at the 2.0 grade level in oral reading. When Data analysis learning disability typi- retested 1-3 years later for this study, both groups had a mean To generate the quantitative reports, digital filtering meth- ial reading skills despite of 13.3 years. Control reading grade levels averaged 11.1 for ods were used to calculate absolute power values for each EEG kction, and conventional silent and 10.6 for oral reading; dyslexics were at 4.9 (silent) derivation for delta (1.5-3.5). theta (3.5-7.5), alpha (7.5-13.5) pie, under study since and 3.3 (oral). Differences in the reading measures are signifi- and beta (13.5-25.0) bands. Relative (percent) power was also :ly dyslexic boys and a cant at P < O.OOO1. calculated for the same 32 lead/band combinations, as well as - 428 C.D. YINGLING ET AL. 1EUROMETRICS AND DY

;-scores for relative power for each lead in each band. These similar results: the false positive rate rose to 5 of 38 controls everely learning disabled thar z-scores express each individual value in terms of the age- (13.2%), while the true positive rate in the dyslexics was 8 of 38 ,pecific LD group ‘had IQ s adjusted mean and standard deviation derived from develop- (21.0%).again not a significant difference (x2= 0.37, P > 0.50). and WRAT standard scores mental equations based on EEG values obtained from several mthmetic slulls.’ By these CI hundred normal children studied by John’s group at NYU. EEG spectra: comparison with local control group [Q who were performing a We first evaluated the data using the most conservative When we applied the same decision rules to the deviance reading quotients) were ckSi criterion for abnormality described by Ahn et al. (1980): ‘the scores based on normative EEG spectral values derived from trast, our dyslexics were all threshold for inferring probable dysfunction (is defined as) our own control group, we found an abnormality rate of 3/38 brlow that expected on the b approximately twice the number of significant values expected (7.9%)in the control group, and 6/38 (15.8%)in the dyslexics. were thus clearly more disab by chance . . . .’ Therefore, they called a given case abnormal if This difference was again not significant ( x2 = 0.50, P > 0.10). However, the neurometric a 2 or more of the 32 spectral values (4 bands X 8 leads) differed ~3slower than that in their from those of the age-adjusted normal population at the P < Evoked potential features cantly different from that 01

0.01 level. We used one-tailed tests of significance since Ahn et Finally, examination of the EP indices provided in the &de that the neurornetnc I al. reported the learning disabled group to have excess relative quantitative reports revealed no features which significantly associated with learning disa delta and theta, and deficient relative alpha and beta power. discriminated between dyslexics and controls. ‘Abnormality’ other factors. We also examined the hit rate at the less conservative 0.05 level rates ranged from zero (signal/noise ratio of visual EP) through The most likely reason to increase the sensitivity. 75% (coherence between homologous lead pairs for visual EPr): findings and those of Ahn e‘ In addition, we used the EEG spectral values from our however, no EP features revealed a low false positive rdte our dyslexics and their LD g control group to define a set of local norms. Using the raw coupled with a high abnormality rate in the dyslexic group. cal and/or sensory deficits power values from the Neurometrics, Inc. quantitative reports, sample but not theirs. Altk we computed a new set of z-scores for each of the 32 values for more carefully screened tha each subject. We then used these z-scores to determine the the similar false positive ral Discussion number of abnormal records in both our groups, using the the two normal groups we, same criterion of 2/32 values differing from the norms at our dyslexics were less neur The first important result from this study is that the fhe P c 0.01. positive rate of EEG spectral abnormalities in our conrrol al specific LD groups. havi In addition to spectral power values derived from eyes criteria for normal develo group normal readers from the San Francisco area .\as closed EEG, the quantitative reports generated by Neuromet- of similar to that reported by Ahn et al. (1980) for norinal vi>ion and hearing status a rics, Inc. also contained several parameters derived from the report did not specify neul subjects from both the New York area and from Barbados. evoked potential data. including signal-to-noise ratios, asymme- This finding suggests that the distribution of EEG specrral specific learning disabled L try, coherence, and reactivity. Evaluation of these data, how- values in all 3 normal groups is similar, and thus that the fact contain many subject5 ever, is complicated by several factors: there are no pub- findings which would have (I) normative values published by John et al. (1980) are an dp- lished data results of the EP analyses; (2) the normative on propriate benchmark against which to assess learning disahled our dyslexic group (E.R. Jc data base was not used to generate deviance scores for the EPs, We conclude that the nc or other patient groups for the presence of quantitative devi- whch were instead evaluated based on a priori criteria derived ances in EEG recorded during resting conditions. to resting EEG spectra, dc from the literature; and (3) no significance levels were provided However, using deviance scores based on the New \‘ark from normal readers when for features noted as abnormal, making it difficult to determine normative spectral values, we found no significant difference in While reading disabilities the relative severity of EP findings. Whle we were thus unable rates of neurometric abnormalities between severely disahled other overt neurologkd fa to formally evaluate the EP data in the same manner as the readers and the normal reading group. We thus failed to can exist without other net EEG spectral data, we did examine these features to determine abnormalities. Although ‘1 replicate the results of Ahn et al. (1980), who reported a high whether any appeared to distinguish the dyslexic and control abnormality rate in their LD groups. to an underlying disorder groups. b! neurometric evaluatior Since our control group was carefully screened to rule out sensory and neurological deficits which may have been preen1 based on recordings obta in the New York normals, it was possible that the New I‘ork tions may prove more sen The 45-50% detection Results norms contained more variance, and would result in a less sensitive test than if norms were based on our controls. How- reflects EEG abnormaliti EEG spectra: comparison with New York normative values ever, when we recalculated the deviance scores using our :lor- and/or sensory dysfuncti This interpretation is I When compared to the New York normative values for the mal readers to provide the normative data, we obtained es\en- re. Ahn et al. that ‘excess ~ EEG power spectrum, 2 of 38 subjects in our control group had tially the same result. The failure to discriminate between aur paneto-occipital regions abnormal neurometric values by the initial criterion of 2 or dyslexic and normal readers is thus not due to the use of an more values deviant at the P < 0.01 level. Thls false positive inappropriate normative group. frequency or regional abr activity is not a very spec rate of 5.3%is comparable to the 4%rate reported by Ahn et al. The most important finding of our study is the discrepdncy with a variety of brain d) (1980). However, only 4 of 38 dyslexic subjects were classified between our EEG spectral findings and those reported by \hn and anoxia (Niedermeyf as abnormal, for a true positive rate of 10.5%, considerably et al. (1980). using identical procedures. If the neurometric study. we found no evid lower than the 47% in the Ahn study. and not significantly abnormalities which they observed were specifically associated other EEG features whic different from the false positive rate in our controls (xz= 0.18, with learning disabilities, our dyslexic group might have been ~tyspecifically associate P > 0.10). expected to produce a higher abnormality rate than the) re- In our previous study, Using the less conservative P < 0.05 criterion produced ported For LD children, since our dyslexic group was more Z.D. Y INGLING ET AL. gEUROMETRlCS AND DYSLEXIA 429

:rose to 5 of 38 controls Werel! learning disabled than their specific LD subjects. Their ferences in delta, theta, or alpha power. However, we did find a 1 the dyslexics was 8 of 38 LD group 'had IQ scores above 85 on the WISC-R, reliable reduction in beta power in the dyslexics (Fein et al. :nce ( x2 = 0.37. P > 0.50). ad WRAT standard mres below 90 in language and/or 1986), which would not have been detected by the neurometrics arithmetic skills.' By these criteria, children with low-average procedure due to differences in the bandwidth of the EEG ontrol group IQ who were performing at expected levels (i.e., 'normal' spectra. (Our findings were primarily in the 19-24 c/sec region sion rules to the deviance reading quotients) were classified as learning disabled. In con- of the beta band and would have been obscured in the neuro- :ctral values derived from trast. our dyslexics were all reading at a level substantially metrics data since the entire 13-25 c/sec beta range was abnormality rate of 3/38 I t~losthat expected on the basis of age and IQ. Our dyslexics banded together.) 18 (15.8%)in the dyslexic, were thus clearly more disabled than their specific LD group. Finally, the evoked potential features provided in the neuro- :ant (x2= 0.50, P > 0.10) HoweLer, the neurometric abnormality rate in our dyslexics metrics quantitative reports also failed to differentiate between was lower than that in their specific LD group and not signifi- our dyslexic and control groups. The wide range of 'abnormal' cantl! different from that of our controls. We therefore con- values obtained for different EP features, even within our ' indices provided in the clude rhat the neurometric abnormalities they found are not control group, indicates that reliance on a priori criteria for :atures whch significantl! associated with learning disabilities per se. but must be due to evaluation of such data is inappropriate. Statistical treatment id controls. 'Abnormalit!' other factors. of these features. with deviance scores calculated with reference ratio of visual EP) through The most likely reason for the discrepancy between our to the normative data base as is done for the EEG spectral s lead pairs for visual EPh I. [Indings and those of Ahn et al. (1980) is a difference between features, may well produce neurometric EP indices specifically a low false positive rate our d\ slexics and their LD groups in the presence of neurologi- sensitive to LD; the ori@nally implemented a priori features .e in the dyslexic group. cal and/or sensory deficits, which were screened out of our are not. While several recent studies (Duffy et al. 1980; John- sample but not theirs. Although our control group was also stone et al. 1984; Shucard et al. 1984) have reported EP morr carefully screened than the New York normative group. differences between normal and dyslexic children, none of the 5imilar false positive rate which we obtained suggests that these studies have employed both large samples and ap- the iwo normal groups were roughly equivalent. In contrast, propriate statistical procedures for quantitatively assessing de- viance. and so must be regarded as tentative. this study is that the fal,e our dyslexics were less neurologically impaired than the Ahn et normalities in our contrid al. specific LD groups, having met the same rigorous screening cnteria for normal developmental hstory and neurological. e San Francisco area wa\ et al. (1980) for normdl vision and hearing status as had our controls. The Ahn et al. report did not specify neurological screening criteria for their ; area and from Barbado,. LA methode neurometrique ne permet pas la detection dons itribution of EEG spectral spec'fic learning disabled group; however, their group did in dyslexies 'pures' similar. and thus that the fact contain many subjects with neurological and/or sensory in et al. (1980) are an ap- findings which would have caused them to be excluded from our dyslexic group (E.R. John, personal communication). Trente-huit garpns profondement dyslexiques et 38 autres I to assess learning disabled sachant bien lire ont et6 testes a I'aide de la methode esence of quantitative de\i- We conclude that the neurometrics procedure, when applied neurometrique. Cette technique de diagnostic est bas& sur ing conditions. io rfsting EEG spectra, does not discriminate severe dyslexics from normal readers when neurological status is held constant. I'application de la taxonomie numerique au spectre EEG ob- 's based on the New Yorh tenu au repos, et complet& par certaines des caracteristiques 1 no significant difference in While reading disabilities may in some cases be secondary to other overt neurological factors, it is clear that severe dyslexia des potentiels hoques. Elle permet d'obtenir des indices de 6 between severely disabled can exist without other neurological symptoms or marked EEG variation pour le spectre EEG, ceci en comparant chacune des group. We thus failed 11) '1980), who reported a high abnormalities. Although 'pure' dyslexia is presumably also due valeurs individuelles a celles obtenues avec une population de to an underlying disorder of brain function, it is not detected reference et a et6 consideree comme permettant de differencier )S. des enfants presentant non des difficult& de lecture (Ahn et irefully screened to rule out bq neurometric evaluation of the resting EEG. Future studies ou .hich may have been present based on recordings obtained under active behavioral condi- al. 1980). Dans notre etude. tous les sujets. dyslexiques et contrhles. possible that the New Yorh tions may prove more sensitive. ont subi des examens tres stricts. permettant de s'assurer de and would result in a lesr The 45-50% detection rate in the Ahn et al. study probably leur normalite intellectuelle, neurologique, sensorielle et lased on our controls. Hou- reflects EEG abnormalities associated with overt neurological emotionnelle. Le taux positif (artificiel) obtenu pour le groupe dance scores using our nor- and/or sensory dysfunction, not with learning disabilities per ive data, we obtained essen- se. This interpretation is further supported by the statement of contrhle a ete comparable a celui deja dkrit. Toutefois. aucun indice de deviation n'a permis de differencier de faGon signifi- to discriminate between our Ahn et al. that 'excess slow waves (delta plus theta) in the parieto-occipital regions were far more frequent than any other cative les dyslexiques des contrbles; la plupart des sujets des IS not due to the use of an frequency or regional abnormality' (1980, p. 1261). Excess slow deux groupes ont ete class& comme normaux. La dyslexie severe en tant que telle n'est donc pas associk a des anomalies our study is the discrepanc) activity is not a very specific sign and has long been associated neurometriques spkifiques, telles qu'elles ont ete dkrites and those reported by Ahn with a variety of brain dysfunctions, including perinatal trauma anterieurement pour une population plus heterogene presentant cedures. If the neurometric and anoxia (Niedermeyer and Lopes da Silva 1982). In this des problemes de lecture. I were specifically associated study, we found no evidence for either excess slow activity or exic group might have been other EEG features which might reflect anomalous brain activ- Supported by NICHD Contract N01-HD-8-2824 and iormality rate than they re- it) specifically associated with severe developmental dyslexia. NINCDS Grant R01-NS17657. ir dyslexic group was more In our previous study, we similarly found no consistent dif- 430 C.D. YINGLING ET AL. and Elsevier Scientific Publisher We thank Jane Rowe and Martha Evans Kiersch for per- John, E.R., Kamel, B.Z., Coming, W.C.,Easton, P., Brown. forming the neurological examinations, Gunilla Haegerstrom- D., Ahn, H., John, T., Harmony, T., Prichep, L., Toro, A,, Portnoy for performing the vision and hearing examinations, Gerson, I., Bartlett, F., Thatcher, R., Kaye, H., Valdes. P. and Ken Nemire for assistance in data collection. We also and Schwartz, E. Neurometrics. Science. 1977, 196: thank E. Roy John. Michael Bergelson, Hansook Ahn, Eileen 1393-1410. SLOW-WAVE OSCl Maisel and the staff of Neurometrics, Inc., for their cooper- John, E.R., Ahn. H., F’richep, L., Trepetin, M., Brown, D. and THE HUMAN BRAE ation and assistance. Kaye, H. Developmental equations for the electroencepha- logram. Science, 1980, 210: 1255-1258. 5.1. MOISEEVA and Z.A. Johnstone, J., Galin, D., Fein, G., Yingling, C.D., Herron. J. Institute of Experimenral k References and Marcus, M. Regional brain activity in dyslexic ind control ctuldren during reading tasks: visual probe evenr-re- (.\ccepted for publication: Ahn, H., Prichep, L., John, E.R., Baird, H., Trepetin, M. and lated potentials. Brain Lang., 1984, 21: 233-254. Kaye, H. Developmental equations reflect brain dysfunc- Niedermeyer, E. and Lopes da Silva, F. Electroencephalogra- tions. Science, 1980, 210: 1259-1262. phy: Basic Principles, Clinical Applications and Related Critchley, M. The Dyslexic Child. Thomas, Springfield, IL, Fields. Urban and Schwarzenberg, Baltimore, MD, 1982. Summary Slow-wavi 1970. Shucard, D.W., Cummins, K.R. and McGee, M.G. Event-re- structures of parkinsonic a Duffy, F.H., Denckla, M.B., Bartels. P.H.. Sandini. G. and lated brain potentials differentiate normal and disahled During slow-wave , th Keissling L. Dyslexia: automated diagnosis by com- readers. Brain Lang., 1984, 21: 318-334. d5 the amount of oscillatil puterized classification of brain electrical activity. Ann. Touwen, B.C.L. and Prechtl, H.F.R. The neurological examtna- odlatory components del Neurol., 1980, 7: 421-428. tion of the chdd with minor nervous dysfunction. Top prevailing. Fein, G.. Galin, D., Yingling, C.D., Johnstone, J., Herron, J. develop. Med., 1970, 30: 1-105. The data obtained rev1 and Davenport, L. EEG spectra in dyslexic and control the hypothetical adaptoge boys during resting conditions. Electroenceph. clin. Neuro- physiol., 1986. 63: 87-97. Keywords: sleep - humar

A number of aut1 nificance of slow : brain potentials for t emotions, memory 1957, 1964, 1979; / 1961; Caspers 1962 1974; Iliukhina et a1 1983). From among th oscillations of brair ing from seconds tc oscillations with pe tract the attention ( lar oscillations occ during changes in t mov 1961; Aladjal We were particu the constant compc develops slow oscil the course of adapt 1977). These osci u ithin the range with the fluctuatio in animals and cardiointervalograi ing adaptive syster

0013-4649/86/$03.50