INTRODUCTION
In recent years EEG biofeedback training has been applied to an increasing number of psychological, neurological, and psychosomatic conditions (e.g., Fleischman, 1997; James & Folen, 1996; Byers, 1995; Tansey, 1993). Sensorimotor (SMR; typically 12-15 Hz) and beta (15-18 Hz) neurofeedback, a form of training designed to enhance intermediate frequency EEG instantaneous amplitudes, has been reported to improve epilepsy (Lantz & Sterman, 1988; Tozzo, Elfner, & May, 1988; Sterman & MacDonald, 1978), attention deficit hyperactivity disorder (ADHD) (Lubar, Swartwood, Swartwood, & O’Donnell, 1995; Rossiter & LaVaque, 1995; Lubar & Shouse, 1976), specific learning disabilities (Tansey, 1985; Linden, Habib, & Radojevic, 1996), and some conditions associated with ADHD such as bruxism, tics, and mood swings (Alhambra, Fowler, & Alhambra, 1995; Tansey, 1986). Minor closed head injury, multiple sclerosis, autism, chronic fatigue syndrome, and pre-menstrual syndrome, head a growing list of conditions reported by clinicians to be partly or fully remediated by SMR-beta neurofeedback training (cf. Othmer, in prep).

The apparent diversity of disorders impacted by SMR-beta neurofeedback training suggests a commonality of mechanisms for these conditions, a fact that should be addressed by any theory that attempts to identify the therapeutic mechanism of SMR-beta neurofeedback. Sterman (1982) proposed that SMR neurofeedback may restore regulated function of thalamocortical mechanisms associated with arousal. In particular, abnormal sensorimotor arousal or excitability may interfere with higher cognitive functions in a resource-limited competive model (Sterman, 1996). Abarbanel (1995) formulated a similar model of self-regulation in which attentional processing were modulated by thalamocortical and limbic circuitry. In his model long-term potentiation was responsible for any functional permanence associated with training. Both models presume SMR-beta neurofeedback impacts functions that modulate arousal (Sterman, 1982; Abarbanel, 1995). Both models readily address the symptomatology and possible mechanisms of ADHD and epilepsy. The primary symptoms of ADHD, inattention, impulsivity, or hyperactivity, are associated with decreased arousal in frontal cortex and subcortical regions (Zametkin et al., 1990; Mann, Lubar, Zimmerman, Miller, & Muenchen, 1991). The cortical hyperexcitability associated with epilepsy may reflect an arousal dyfunction, possibly due to a loss of integrity in the thalamic gating mechanism (Sterman, 1982). In addition to motor or vocal tics, sufferers of Tourette’s Syndrome often exhibit somnambulism, night terrors, and other disorders of arousal (Barabas, Matthews, & Ferrari, 1984). Attentional processes in particular appear to be uniquely sensitive to problems of arousal, and thus they serve to be a good measure of effectiveness in restoring such functions.

The Test of Variables of Attention (TOVA) is a continuous performance task that assesses attentional processes relative to a normative database. The TOVA provides a quantifiable measure of effectiveness of SMR and beta biofeedback training for improving specific attentional properties such as impulse control and response consistency. The lack of test-retest practice effects, the use of language-independent nonverbal stimuli, and an extended test length (22.5 min), all make the TOVA especially useful in evaluating treatment effects in an ADHD, learning disabled, or like population (Greenberg, 1987).

The purpose of the present study is to evaluate the efficacy of SMR-beta neurofeedback for children and adults suffering from attentional problems as measured by the TOVA.

David A. Kaiser and Siegfried Othmer
EEG Spectrum, Inc. Encino, CA
November 1997

Subjects
Four hundred and eight children and adolescents (age 6 to 16 years, mean 10.7) and 122 adults (17 to 67 years, mean 37.2) participated in this study. Females comprised less than one-quarter of the child and adolescents group and nearly one-half of the adults (92 and 58, respectively). Subjects were obtained in nine clinical settings affiliated with EEG Spectrum, Inc. and were selected based on the availability of pre- and post-training data for the TOVA. None of these subjects were on any stimulant or antidepressant medications during the test. Although a plurality of subjects suffered from ADHD, many also exhibited comorbid conditions of more severe behavioral disorders (ODD and Conduct Disorder), Tourette’s Syndrome, minor traumatic brain injury, epilepsy, anxiety disorders, and depression. The subjects also included some who were referred for ADHD but may not have met the classical diagnostic criteria for the condition. Adults varied on diagnosis with the majority exhibiting some form of ADD. Materials
EEG biofeedback training was performed on Neurocybernetics 2-Channel EEG systems. All subjects were evaluated with the Test of Variables of Attention (TOVA) (Greenberg, 1987), a continuous performance task (CPT) that presents to a subject a geometric target or non-target. The use of a single non-target allows this test to be conceptualized as a Go/No-Go task, a form of test which is associated with frontal lobe function (e.g., Levin et al., 1991). Results from the TOVA include measures of omission errors (inattention), commission errors (impulsivity), response time (speed of information processing), and response time variability (consistency of response). Scores are presented in standard scores with every standard deviation presented as 15 points above or below the mean. This test was administered on a PC computer and used a single switch for response. This test consists of only two non-verbal stimuli which requires a subject to pay attention for 22.5 min without prolonged rest. Presentation probabilities for targets and non-targets are mixed between test halves in order to evaluate high-likelihood and low- likelihood response conditions (i.e., 20% targets first half of test, 80% targets second half), and thereby provide measures of impulsivity and inattention, respectively. Normative age-based data is available for each gender; for children, in single-year age groups, and for adults in 10-year age groups (Greenberg & Waldman, 1993).

Procedure
The training protocol consisted of rewarding enhanced EEG amplitudes in the 12-18 Hz frequency regime, while simultaneously inhibiting excessive amplitudes in the low frequency (4-7 Hz) and high-frequency (22-30 Hz) regimes. Electrode placement always included one electrode site on the sensorimotor strip (at either C3 or C4 in the standard 10-20 system) and less commonly one electrode with either frontal or parietal placement. If training was done solely at C3 and C4, then the montage was referential to the proximate ear. If training involved frontal or parietal placement, the montage was bipolar with either C3-Fpz or C4-Pz. Left-side (C3) and right-side (C4) training involved rewarding activity in the 15-18 Hz and 12-15 Hz, respectively. Occasionally, these two protocols were used in succession during a single training session with the respective duration (e.g., 10 min SMR, 20 min Beta) of the two protocols titrated on the basis of changing symptomatology and TOVA results (Greenberg, 1987). Left-hemisphere training (e.g., C3) involved Beta reward only whereas right-hemisphere training involved SMR reward only.

Training consisted of 30 min of visual and auditory feedback on the instrument, within a 45-min contact hour. Visual feedback was provided by a variety of means which map the EEG amplitude in the reward and inhibit bands into the brightness, size, and/or velocity of objects on a computer monitor. Most commonly, information about the amplitude of signals in each of the bands was given independently. Alternatively, the subject was simply be notified that an inhibit threshold was exceeded by the withholding of the conventional reward. When all reward conditions were satisfied for a minimum of 0.5 s, an auditory beep and visual incentive (e.g., highway stripe, star in sky) was provided as reinforcement. The visual feedback signal was occasionally complemented with direct tactile and auditory feedback of EEG amplitude in the reward band.

Subjects were evaluated prior to training and after approximately 20 sessions. Those subjects who were further treated were retested after approximately 40 training sessions. Most subjects completed or discontinued training after 20 sessions (mean 24.1, range 18 to 61 sessions).

A Huynh-Feldt correction for degrees of freedom was applied to all interactions to counter potential nonsphericity of the four dependent measures. When an interaction of condition (treatment by dependent measures) was significant at the .01 level, planned comparison t-tests were used to evaluate differences for each dependent measure. The Bonferroni correction for multiple tests was consequently applied to planned comparisons.

RESULTS
Repeated measures analyses of variance (ANOVA) were used to evaluate the effect of group membership for three factors: medication, gender, and age. Medication refers to whether subjects took condition-specific medication at any time during their EEG biofeedback training. Medication information was only present for 324 subjects and only the data from these individuals were analyzed. No effect of medication, F (2,551)=1.884, ns ; gender, F (2,417)=0.949, ns ; nor age F (2,447)=3.754, p >.01; was found on the TOVA measures. As no significant differences were found between groups, all groups were combined into a single group. Repeated measures ANOVAs were then used to evaluate the effect of EEG biofeedback training on four dependent measures of the TOVA: Inattention (percent omission), Impulsivity (percent commission), Response Time, and Response Variability. Low scores were truncated at four standard deviations below normal (i.e., 40 points). Mean pre- and post-training TOVA scores are presented in Table 1.Table 1.
Mean standard scores for TOVA measures before and after approximately 20 EEG biofeedback sessions for 324 subjects with attentional deficits .