EFFECTS OF PHONETIC PROCESSING AND STIMULUS RELEVANCE ON THE AUDITORY EVOKED RESPONSE By DENNIS ALFRED SILVA A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1977 Copyright by Dennis Alfred Silva 1977 ACKNOWLEDGMENTS I wish to express my sincere appreciation to William Yost, chairman of my committee, not only for his helpful suggestions for the present investigation, but also for his tremendous contribu- tions to my graduate education. In addition, I wish to thank Donald Teas for the use of his laboratory facilities which made the present research possible, as well as for his useful comments on the work itself. I would like to express my appreciation also to my other committee members, Keith Berg, Howard Rothman, Paul Satz, and Robert Isaacson, for their knowledgeable comments and suggestions . ACKNOWLEDGMENTS . LIST OF TABLES . . LIST OF FIGURES ABSTRACT , TABLE OF CONTENTS Page . . iii VI DISCUSSION. VII INTRODUCTION 1 22 Summary 24 METHOD 24 Subjects 24 Stimuli and Apparatus 27 Procedure RESULTS 3 2 Topographical Effects Phonetic versus Nonphonetic Tape Effects 44 Effects of Stimulus Relevance 55 63 Topographical Effects 64 Stimulus Relevance 71 APPENDIX A: Ordering of Stimuli on Tape 1 74 APPENDIX A: Ordering of Stimuli on Tape 2 75 APPENDIX B: Order of Experimental Conditions 76 APPENDIX C: Summary of Partial Amplitude Data (Microvolts) ■ 77 for Two Additional Subjects REFERENCES BIOGRAPHICAL SKETCH LIST OF TABLES Table Page 1 Summary of Experimental Conditions 29 2 Mean Amplitudes (microvolts) for Major Com- ponents of Evoked Responses Recorded from Three Locations 37 3 Number of Persons Having Greater Left Hemi- sphere Amplitudes for Various Components 39 4 Mean Latencies (msec) for Major Components of Evoked Responses Recorded from Three Locations. . 43 5 Mean Amplitudes (microvolts) for Major Com- ponents of Potentials Evoked by Stimuli on Phonetic and Nonphonetic Tapes 45 6 Mean Latencies (msec) for Major Components of Potentials Evoked by Stimuli on Phonetic and Nonphonetic Tapes 48 7 Significant Wilcoxen Tests between Potentials Evoked by Stimuli on Phonetic and Nonphonetic Tapes 53 8 Mean Amplitudes (microvolts) of Evoked Poten- tial Components in Conditions of Stimulus Relevance 56 LIST OF FIGURES Figure Latencies for N]_ component of potentials evoked by CV-, during phonetic and nonphonetic processing tasks Averaged evoked potentials recorded from left and right hemispheres during phonetic and nonphonetic tasks Averaged evoked potentials recorded from the vertex to CV, and T, evoking stimuli during conditions of stimulus relevance Averaged evoked potentials recorded from Subjects 1-4 during conditions of stimulus relevance Averaged evoked potentials recorded from Subjects 5-8 during conditions of stimulus relevance Page 1 Averaged evoked potentials to CV stimulus recorded from vertex, left, and right temporo-parietal sites during various conditions 34 2 Averaged evoked potentials to T-^ stimulus recorded from vertex, left, and right temporo-parietal sites during various conditions 35 3 R-values and hemispheric voltage differences for all subjects across all conditions 41 50 52 6n 61 62 Abstract of Dissertation Presented to the Graduate Council of the University of Florida in Partial Fulfillment of the Require- ments for the Degree of Doctor of Philosophy EFFECTS OF PHONETIC PROCESSING AND STIMULUS RELEVANCE ON THE AUDITORY EVOKED RESPONSE By Dennis Alfred Silva March 1977 Chairman: William A. Yost Major Department: Psychology There is a substantial body of research data, gathered from both clinical and nonclinical behavioral investigations, which sup- port the phenomenon of specialization of function in the human brain. In addition, recent findings suggest that there are structural asymmetries in the cortex which parallel the functional differences. Results of recent electrophysiological investigations suggest that the left hemisphere's specialization for language is reflected in auditory evoked responses to speech and nonspeech stimuli and during phonetic and nonphonetic processing of speech stimuli. Because of several inconsistencies in the existing data, however, the precise nature of the effect as well as the actual processing mechan- isms reflected in the reported asymmetries remain obscure. For example, the evoked potential asymmetries might be reflecting the activity of a lateralized neural center responsible for phonetic analysis of speech stimuli or they might be due to a lateralized cortical activation occurring in preparation for phonetic processing. One purpose of the present investigation was to investigate further the evoked potential correlates of hemispheric specializa- tion for language by recording auditory evoked responses to a speech and a nonspeech stimulus from left and right temporo- parietal locations (referred to linked mastoids) . Eight subjects participated in a series of phonetic and nonphonetic discrimination tasks. During each condition two speech (CV syllables) and two nonspeech (tones) stimuli were pre- sented to subjects and evoked potentials were averaged to one of each class of stimuli. In one type of task the speech stimuli were phonetically different (/ba/ and /da/) ; in the other type they were phonetically similar (/ba/) but differed in fundamental frequency. By presenting both speech and nonspeech stimuli within a single run while subjects are engaged in tasks reauiring either a "phonetic" or "nonphonetic" cognitive set, it is possible to determine if an evoked potential asymmetry is due to a lateralized preparatory activation or to the activity of an actual phonetic analyzer localized in the left hemisphere. If the former explana- tion is correct, then asymmetrical cortical responses should be manifest in the evoked potential to a stimulus whenever a subject expects to make a phonetic discrimination, regardless of the actual nature of the stimulus presented. If the latter explanation is correct then an asymmetrical response should occur only to a speech stimulus and not to a nonspeech stimulus in spite of a "phonetic set." Analysis of the data demonstrated no significant differences between the left and right hemispheres for mean amplitudes or mean latencies of the evoked potential components. The majority of subjects, however, did have slightly larger left hemisphere responses to both evoking stimuli. Regarding potentials obtained during phonetic and nonphonetic processing of stimuli, the data confirm a previously reported finding of a differential left hemisphere response to a speech stimulus during phonetic processing and extend this observation to the use of natural speech syllables. No differences in potentials occurred to the nonspeech stimulus even when subjects were engaged in the phonetic processing task. This suggests that the dif- ferential left hemisphere response is not simply one of a prepara- tory nature but rather it may reflect the activity of a phonemic analyzer. The possibility of two left hemisphere mechanisms was discussed — one for speech detection and one for phonetic analysis. Finally, vertex potentials were analyzed for effects of stimulus relevance, with the stimulus which subjects were discrimi- nating being the relevant stimulus and the other three stimuli being irrelevant stimuli. It was found that the P3 component of the evoked potential was most sensitive to various conditions of stimulus relevance. The amplitude of the P3 component was largest when the evoking stimulus was most like the relevant stimulus, and smallest when the evoking stimulus was of the other class (i.e. speech or nonspeech) than the relevant stimulus. These findings support a "neural template" model of P3. INTRODUCTION Since the nineteenth century when anatomists and neurologists first noted a relationship between sensory-motor disorders of the right half of the body and speech disturbances, mounting scientific evidence attests to the functional disparity in humans between the two cerebral hemispheres with respect to certain cognitive capacities. Over the years the understanding of cerebral specialization in the human brain has grown, and investigation of the phenomenon has expanded from predominantly clinical observations to research with non-brain damaged persons in a variety of experimental paradigms. Novel approaches and more sophisticated apparatus have resulted in an expansion of methodologies for investigating functional locali- zation in the brain. Not only have behavioral experiments flour- ished, but, during the last eight years or so, reports of electro- physiological investigations of the phenomenon have begun to appear in the literature. The purpose of the present report is to expand further our knowledge of evoked potential correlates of language localization in the brain while attempting to integrate some of the existing data on hemispheric asymmetries. The effects of phonetic processing and stimulus relevance on the auditory evoked response to speech and nonspeech stimuli are addressed in the present investigation. With respect to their localization in the brain, speech and language functions are perhaps the most extensively studied and documented of human cognitive processes. A substantial amount of the clinical evidence for the localization of language in the left hemisphere comes from the work of Penfield and Roberts (1959) . Those authors reported that, excluding persons with cerebral injury occurring prior to age two, over 70 percent of their patients who had surgery on the left hemisphere showed transient dysphasia. This was so for both right and left handers. In contrast, they found that fewer than 1 percent of right and fewer than 7 percent of left handers showed dysphasic symptoms following operation on the right hemisphere. Penfield and Roberts (1959) showed also that electrical stimula- tion of certain areas of the left hemisphere can cause an arrest or a hesitation of speech or can result in a variety of dysphasic responses such as misnaming, word distortion or repetition, and perseveration. They reported that stimulation of three regions of the brain in par- ticular could alter speech processes: (1) Broca's area (the three gyri anterior to the lower precentral gyrus) ; (2) the supplementary motor area (on the medial surface of the brain anterior to the precentral leg area) ; and (3) the posterior temporo-parietal area (the posterior part of the first, second, and third temporal convolu- tions; the supramarginal gyrus; and the angular gyrus) . Lateralized cortical lesions have been shown to result in significant deficits in performance on verbal intelligence tests when these lesions occur on the left side but not on the right (Heilbrun, 1956; Klove, 1959; McFie & Piercy, 1952; Reitan, 1955; Satz, Richard, & Daniels, 1967), although Milner (1958) noted that if the disorder is restricted to the left temporal lobe this deficit does not occur. She did report, however, as had others previously, that learning capacities and memory for verbal material is impaired following left temporal lobe lesion (Meyer & Yates, 1955; Milner, 1958) . Additional clinical evidence for lateralization of language functions in the left hemisphere comes from persons having undergone the Wada sodium amytal test (Branch, Milner, & Rasmussen, 1964; Wada & Rasmussen, 1960) or the surgical sectioning of the forebrain commissures. The Wada test was developed as an alternative to using manifest handedness as the sole means for identifying cerebral dominance in patients about to have surgery. The technique consists of injecting a barbiturate, sodium amytal, into the right and left carotid arteries on separate occasions, thereby influencing the functioning of the right and left hemispheres individually. The Wada treatment results in a temporary interference with normal hemisphere functions and thus allows for comparison of the roles of each side of the brain in speech and language. Treatment on the side of the language domi- nant hemisphere, therefore, typically results in a transient inter- ference of expressive and receptive language functions (Branch et al . , 1964; Perria, Rosadini, & Rossi, 1961; Terzian, 1964; Wada & Rasmussen, 1960) . In contrast, injection of the barbiturate on the nondominant side has been reported to impair musical ability (Bogen s Gordon, 1971; Gordon & Bogen, 1974). Sectioning of the forebrain commissures for the control of epileptic seizures has provided a yet different means to evaluate the capabilities of each hemisphere, independent of the influence of the other. In the human, commissural sectioning has been success- ful clinically, but as a research resource human studies have been somewhat inconclusive. In most of these cases the left hemisphere is clearly the dominant one for language functions although there is some evidence that the right hemisphere does possess at least minimal receptive and expressive capabilities (Gazzaniga & Hillyard, 1971; Gazzaniga & Sperry, 1967; Levy, Nebes, & Sperry, 1971) . Because of factors such as preexisting brain damage and cross-cueing strategies between the two hemispheres, however, more sophisticated capabilities of the nondominant right hemisphere remain unresolved. Over the years, investigators have searched for structural and physiological differences between the hemispheres that might account for observed differences in function. Recently, several studies have reported left-right asymmetries in humans in the region posterior to Heschl's gyrus on the superior aspect of the temporal lobe (Geschwind & Levitsky, 1968; Wada, Clarke, & Hamm, 1975; Witelson & Pallie, 1973) . This area, the planum temporale, comprises 5 part of the region classically known as Wernicke's area and has been shown to be of major importance in language function (Penfield & Roberts, 1959) . This left-right anatomical difference has been reported to be present as early as the 29th gestational week (Wada et al., 1975). Other asymmetries have been reported to exist in humans: the length of the Sylvian fissure, with the left being approximately 10 mm longer than the right (Yeni-Komshian & Benson, 1976) ; and the pathway of the middle cerebral artery in the region of the Sylvian fissure which suggests a larger parietal operculum on the left (LeMay & Culebras, 1972) . In addition to the research with clinical populations described earlier, a vast number of behavioral investigations have been con- ducted with nonclinical subjects. Experimental paradigms in both the visual and auditory modalities have provided a means for investi- gating laterality of function in the cerebral hemispheres. In man approximately half of the fibers originating from each eye are contralateral. Fibers from the nasal half of the retina cross at the optic chiasm and higher order neurons eventually termi- nate in the opposite striate cortex. Fibers from the temporal halves remain uncrossed and their higher order neurons eventually terminate in the ipsilateral receiving area. Consequently, percep- tion of visual stimuli in the left visual half-field (impinging on the right hemi-retinae) occurs in the right occipital lobe, and perception of visual stimuli in the right visual half-field (imping- ing on the left hemi-retinae) occurs in the left hemisphere. 6 Experimenters have found that when a stimulus is presented to an observer so that it is confined to a single visual half-field (i.e. the left or right), the field of presentation which results in the superior recognition is dependent on the nature of the stimulus. For example, a right half-field superiority is obtained most frequently with words (Kaufer, Morais, & Bertelson, 1975; McKeever & Huling, 1971; Mishkin & Forgays, 1952; Terrace, 1959), letters (Bryden, 1965, 1966; Bryden & Rainey, 1963; Heron, 1957; Kimura, 1966), and digits (Hines & Satz, 1971; White, 1969), whereas no signficant field differences are usually found when stimuli consist of nonalphabetic material such as geometric shapes or nonsense forms (Bryden, 1960; Bryden & Rainey, 1963; Heron, 1957; Kimura, 1966; Terrace, 1959) . Recognition of faces is superior when they are presented to the left visual half-field (Ellis & Shepherd, 1975) . Because words and letters, but not geometric or nonsense forms, are identified more accurately in the right visual field, Kimura (1966) suggested that the more direct pathway from that field to the language dominant left hemisphere is the basis for the visual field differences. Reaction time data employing the visual half-field paradigm also support a cerebral dominance model: observers respond faster to verbally coded stimuli when they are presented in the right visual half-field (Cohen, 1972; Geffen, Bradshaw, & Nettleson, 1972; Geffen, Bradshaw, & Wallace, 1971; Seamon & Gazzaniga, 1973) , while the use of nonverbal stimuli results in faster reaction lines when they are presented in the left visual half-field (Rizzolatti, Umilta, & Berlucci, 1971) . In the auditory modality a dichotic listening paradigm has been employed to study hemispheric differences in the processing of stimuli. This technique, introduced by Broadbent (1954), con- sists of presenting conflicting stimuli simultaneously to the two ears. The listener's task is to report the stimuli which he perceives. Using the dichotic technique, Kimura (1961) found that both normal and temporal lobe brain damaged persons usually had higher scores for reporting verbal material presented to the right ear than to the left ear, with the exceptions being persons who had speech functions localized in the right hemisphere as determined by the Wada test. Since the left hemisphere is the language dominant hemisphere in the vast majority of persons (Penfield & Roberts, 1959) the ear asymmetry which occurs with dichotic stimulation is usually referred to as a right ear advantage. Since Kimura ' s early studies, the right ear advantage has been replicated with a variety of verbal stimuli including meaningful and nonsense words (Curry, 1967; Curry & Rutherford, 1967); digits (Broadbent & Gregory, 1964; Dirks, 1964) ; and backwards speech (Kimura & Folb, 1968) . Shankweiler and Studdert-Kennedy (1967) have shown that a right ear advantage is obtained with consonants but not with vowels. To explain her findings, Kumura suggested that the crossed audi- tory pathway in man is more efficient than the uncrossed pathway, thus providing the ear contralateral to the language hemisphere with a superior channel to the speech processor. Unlike the visual path- ways, the fibers in the auditory system are bilateral. By proposing a greater efficiency for the contralateral pathway and a suppres- sion of ipsilateral fibers during dichotic stimulation, however, Kimura (1967) was able to interpret the right ear advantage in audition in the same vein as the perceptual asymmetries in visual half-field investigations, i.e. superior transmission to the language dominant left hemisphere. Support for her model is derived from electrophysiological recordings in lower mammals which suggest that the contralateral auditory pathways are superior to the ipsilateral pathways (Benson & Teas, 1976; Hall & Goldstein, 1968; Rosenzweig, 1951; Tunturi, 1946) . Kinsbourne (1970) has proposed a model which attempts to account for laterality differences in perceptual experiments on the basis of an attentional bias toward stimuli coming from either the left or the right rather than on a perceptual asymmetry resulting from differences in the neural pathways from sensory receptors to the language process- ing hemisphere. He suggested that the use of verbal stimuli "induces preparatory left hemisphere activation, and thus biases attention" (Kinsbourne, 1970, p. 196) to stimuli on the contralateral right side. Because of this selective attention to the right side, verbal stimuli to the right ear are more accurately recognized than verbal stimuli to the left ear. Likewise, since the right hemisphere is special- ized to process nonverbal stimuli, their use activates the right hemisphere which in turn facilitates perception of stimuli presented to the contralateral left side. To support his model Kinsbourne presented data which showed that subvocal rehearsal of verbal stimuli (a process which should activate the language dominant left hemisphere) results in an increased perception of stimuli on the right side but not on the left. At present it remains to be determined whether Kimura ' s perceptual asymmetry model or Kinsbourne 's attention model is the better one, since both can account for the majority of the existing data. Both, however, rely on the left hemisphere's specialization for speech and language and thus support a cerebral dominance model for explaining performance asymmetries in visual and auditory experi- ments. In the early 1970 's researchers began to investigate the possi- bility that lateralization of function in the cerebral hemispheres is reflected in cortical potentials recorded from the human scalp. Some researchers looked at ongoing electroencephalograms (EEG's) and found relatively less alpha activity on the left side when subjects were engaged in linguistic tasks (Galin & Ellis, 1975; Galin s Ornstein, 1972; McKee, Humphrey, & McAdam, 1973; Morgan, MacDonald, s, Hilgard, 1974; Robbins & McAdam, 1974; Wilson, Vieth, 10 & Darrow, 1957) . in contrast, if a task is spatial in nature, a greater decrease in alpha is seen over the right hemisphere (Galin & Ellis, 1975; Galin & Ornstein, 1972; Morgan et al., 1974; Morgan, McDonald, & MacDonald, 1971; Robbins & McAdam, 1974). A second electrophysiological approach to studying hemispheric processing of information has been to compare sensory evoked cortical potentials recorded from the left and right hemispheres to linguistic and nonlinguistic stimuli. The sensory evoked response reflects electrical activity of the brain which is time-locked to a change in sensory input. When recorded from the human scalp the evoked response is typically represented as an average of many responses so as to minimize the random background activity of the EEC In the auditory modality, the evoked response waveform has been sub- divided into early components which occur within 8 msec, middle latency components occurring between 8 and 40 msec, and late com- ponents with latencies greater than 40 msec (Picton, Hillyard, Krausz, & Galambos, 1974) . It is the late components, reflecting cortical activity, which have been investigated with respect to hemispheric differences. When recorded from the vertex the late potentials include a series of alternating positive and negative waves traditionally labelled as P-^ N-^ and V 2, with latencies of approximately 50, 100, and 180 msec, respectively (Davis, Mast, Yoshie, & Zerlin, 1966; Davis & Zerlin, 1966; Rothman, Davis, & Hay, 1970) . 11 Depending on the experimental conditions, a third positive com- ponent may occur, typically at 300 msec or longer following stimulus onset (Sutton, Braren, Zubin, & John, 1965; Sutton, Tueting, Zubin, & John, 1967) . This late positive wave is usually referred to as P3 or, because of latency, P300. The negative peak between P2 and P3 is called N2 and usually appears between 250 and 300 msec post- stimulus onset. Although the N-j_ and P2 components can be affected by changes in the stimulus parameters of intensity and duration (Butler, Keidel, & Spreng, 1969; Davis et al . , 1966; Onishi & Davis, 1968; Rapin, Schimmel, Tourk, Krasnegor, & Pollack, 1966), they are also sensitive to certain psychological variables. For example, the Nti-P- peak- to-peak amplitude is increased when attention is enhanced through counting (Gross, Begleiter, Tobin, & Kissin, 1965; Mast & Watson, 1968) and in discrimination/detection tasks (Davis, 1964; Hirsch, 1971) . When subjects are told to listen selectively to stimuli arriving through a particular channel (e.g., sense modality), the N1-P2 amplitude is larger than the corresponding amplitude evoked by the same stimuli arriving through a nonattended channel (Debecker S Desmedt, 1966; Satterfield, 1965; Spong, Haider, & Lindsley, 1965). Naatanen (1975) has chosen to explain this effect as being due to an increase in a subject's general arousal as opposed to his selective attention, but recent data presented by Schwent and Hillyard (1975) conflict with Naatanen interpretation. 12 By presenting stimuli rapidly through different auditory "channels" (i.e., lateralized to the left or right) they were able to keep the subjects' arousal levels from fluctuating from stimulus presen- tation to stimulus presentation. The P3 component of the evoked potential was shown by Sutton and colleagues to be dependent on psychological variables (Sutton et al., 1965, 1967). Since then, several investigations have been conducted in an effort to deliniate precisely what these variables are. It has been shown that the P3 wave can be evoked by stimuli which deliver information relevant to the task of a subject (Donchin & Cohen, 1967; Sutton et al., 1965, 1967), by stimuli about which a decision must be made (Rohrbaugh, Donchin, & Eriksen, 1974; Smith, Donchin, Cohen, & Starr, 1970) , and by novel or unexpected stimuli (Ritter & Vaughan, 1969; Ritter, Vaughan, & Costa, 1968) . In 1971 reports began to appear in the literature describing hemispheric differences in auditory evoked responses to speech and nonspeech stimuli. Unfortunately, it is not possible to integrate all the data presently available because neither methodologies nor dependent measures have been consistent among different investi- gators. Morrell and Salamy (1971) measured responses from frontal leads (over Broca's area), from Rolandic leads (over the sensory-motor field of vocalization muscles) , and from temporo-parietal leads 13 (over Wernicke's area) to verbal nonsense words. They compared the peak amplitudes of the N, and P2 components and the peak-to- peak N1~P2 amplitude. They found that the N, wave was signifi- cantly larger from the left hemisphere than from the right hemis- phere, particularly in the temporoparietal region. No hemisphere difference was found when the amplitudes of the P2 waves were compared. In addition, those authors reported that the peak-to- peak measurements tended to obscure the hemisphere differences that did occur. Since nonverbal stimuli were not employed in that study it is not known whether the hemispheric asymmetries obtained by Morrell and Salamy were specific to the use of speech or speech-like signals or if they might have occurred also with other stimuli. Cohn (1971) compared the left and right hemisphere responses to monosyllabic speech stimuli and to click stimuli. He reported that in about half of his subjects the verbal stimuli produced an initial negative wave, largest in the temporo-parietal region, which peaked between 30 and 50 msec. The remaining subjects showed no hemisphere differences with these stimuli. The signifi- cance of Cohn's findings is questionable, however, because the stimuli he employed were not equated on the basis of duration, rise-decay time, frequency composition, and other physical parameters. Molfese, Freeman, and Palermo (1975) presented noise stimuli, nonsense syllables, and monosyllabic words to babies, children, 14 and adults. In all three groups they found greater left hemi- sphere Nj-Pj amplitudes for both classes of verbal stimuli and greater right hemisphere N-i-P2 amplitudes for the noise stimuli. Studies looking at hemispheric differences to speech and nonspeech stimuli when subjects were actively engaged in discrimi- nation or detection tasks have been reported also. Friedman, Simson, Ritter, and Rapin (1975a) averaged potentials to human speech and human sounds when they comprised a "no task" condition, when they were "signal," and when they were "nonsignal" stimuli in a vigilance task. They reported that the amplitude of the left hemisphere N-j_ component of the evoked response to signal words was greater than the corresponding wave recorded from the right hemisphere. The asymmetry did not occur to signal sounds. No differences in N^ or P2 waves were obtained in the nonsignal or the no task conditions. Those authors did not report peak-to-peak Nl-P2 amP1;'-tudes- In a second study, Friedman and his colleagues found larger left hemisphere N^ amplitudes to words which delivered task-relevant phonemic information (Friedman, Simson, Ritter, & Rapin, 1975b). In both of Friedman's studies hemispheric asymme- tries in the P^ waveform were inconsistent. Matsumiya, Tagliasco, Lombroso, and Goodglass (1972) investi- gated the effect of the "meaningfulness" or "significance" of verbal and nonverbal stimuli to determine if this factor, rather than the verbal nature of the stimuli per se, could be responsible 15 for the differential hemisphere response. As a measure of asymmetry they employed the ratio R = W, / (W. + W ) where "W" refers to the peak-to-peak amplitude of the N.-P- components and "1" and "r" refer to the left and right hemisphere responses, respectively. They reported that the response to meaningful words showed the greatest asymmetry, followed by that to meaningful noverbal stimuli. The latter, in turn, was larger than the asymmetry occurring in the two nonmeaningful stimulus conditions. Even though greater asymmetries were observed during the meaningful conditions than in the others, most subjects did show a larger left hemisphere response in all conditions. Galambos, Benson, Smith, Schulman-Galambos and Osier (1975) recently investigated the auditory evoked response to speech (/ba/; /pa/) and tonal (250 Hz; 600 Hz) stimuli and found no significant hemispheric differences in the mean amplitudes of the major com- ponents. In analyzing the output voltage at each digitized point along the time dimension of the waveforms, the general shape of the left hemisphere's response to the speech stimuli was found to differ substantially from that same hemisphere's response to the tones, particularly in the region of the N-^ and P3 components. Those authors noted that although the group differences were not par- ticularly large, some of the subjects had marked hemispheric asymmetries. The right hemisphere's responses to the speech and tone stimuli, however, were found to be quite similar to each other. 16 Wood, Goff, and Day (1971) designed an experiment that allowed them to compare each hemisphere's response to a speech- like signal when the signal was part of a "linguistic" and a "nonlinguistic" discrimination task. In the former condition a subject had to make a phonemic discrimination between /ba/ and /da/,- in the latter the stimuli differed in fundamental frequency, and thus required only a pitch discrimination (/ba/-low versus /ba/-high) . The stimuli were such that /ba/ in the linguistic task was identical to the /ba/-low in the nonlinguistic discrimina- tion task. Their data analysis consisted of a point by point comparison of the digitized output as described above. They found no difference between the shapes of the right hemisphere's responses to the common /ba/ in the two tasks, but the response obtained from the left side when subjects made the phonetic discrimination was significantly different at several time points from that hemisphere's response when only a pitch discrimination was made. Those authors suggested, therefore, that the left hemi- sphere responds differently to a stimulus during linguistic analysis than it does during nonlinguistic processing. In 1975, Wood replicated the findings of his earlier study using different steady state formants and phonemes with different places of articulation. In addition, he included a condition in which subjects had to discriminate between two second formants iso- lated from acoustically different, but linguistucally similar stimuli. Since he found no differential left hemispheric response 17 to the formants, he concluded that the previous finding with actual speech stimuli was related to the processing of the linguistic aspect of the stimuli and not simply the acoustic processing of the formant transitions of the stop consonants. At the present time there remain several inconsistencies in the existing data on hemispheric asymmetries of auditory evoked potentials. Because of this, the cortical functions which under- lie these asymmetries remain obscure. The fact that some investi- gators (e.g. Molfese et al., 1975) have observed a hemispheric asymmetry in the amplitudes of the evoked potential components when subjects were passive participants in an experiment suggests that the differential left hemisphere activity might reflect a physiological "linguistic detector." This interpretation would be consistent with the report of hemispheric asymmetries in babies (Molfese et al., 1975). The existence of such a detector which responds to phonemic transitions in a speech signal would explain the fact that asymmetries occur as early as 100 or 200 msec following stimulus onset, when much of the stimulus would not yet have been heard. Other data, however, are inconsistent with the notion of a lateralized linguistic detector. For example, if such a device does exist then Friedman and his colleagues (Friedman et al., 1975a) should have observed significant hemispheric asymmetries to words when they were of a nonsignal nature and when they com- prised the no task condition just as they did when the words were signal stimuli. 18 Galambos et al. (197 5) showed that the waveforms recorded from the left hemisphere in response to speech and tonal stimuli were different while those recorded from the right hemisphere were not. That finding by itself supports a model for a lateralized detector in the left hemisphere. What is inconsistent, however, is that the asymmetry reported by Molfese et al . (1975) was mani- fest in the amplitudes of the major components (N -P2 amplitude) while in the Galambos study the hemispheric differences were not in amplitude measurements but in the differential shapes of their responses to speech and nonspeech stimuli. In Wood's investigations (Wood, 1975; Wood et al., 1971), hemispheric responses were not compared for speech and nonspeech stimuli. Because of this, the data can be used neither to support nor refute the existence of a lateralized phonemic detector. The left hemisphere's response to a speech stimulus did differ, however, between phoneme discrimination and pitch discrimination. The data do suggest, therefore, the possibility of a lateralized center for phonemic discrimination as opposed to simple phonemic detection. For an alternative explanation of Wood's data, one can borrow from Kinsbourne's (1970) model of selective attention. If, as Kinsbourne suggests, perceptual asymmetries result from a prepara- tory activation of the hemisphere responsible for analysis of an expected stimulus, then most of the evoked potential asymmetry 19 data are confounded with this factor. For example, Friedman et al. (1975) , employing human words and human sounds, presented these two types of stimuli to the subject in separate runs. Because the subjects were aware of the type of processing demanded by the task, differential activation of the hemispheres might have occurred. Similarly, Galambos and his colleagues employed speech and tones in their experiment but these, also, were presented in separate listings (Galambos et al., 1975). The present investigation is designed to investigate further the phenomenon of evoked potential asymmetries as related to the left hemisphere's functional specialization for language processes. In addition, the present set of experiments is designed to investi- gate the role of expectancy or "cognitive set" on the auditory evoked potentials recorded from the two hemispheres. This can be done by presenting both speech and nonspeech stimuli within a single run while subjects are engaged in various tasks demanding phonetic and nonphonetic processing. If a lateralized cortical activation in preparation for verbal analysis of a stimulus does occur, then this effect should be manifest in the evoked response to a stimulus whenever a subject expects to make a phonetic discrimi- nation, regardless of the actual nature of the stimulus presented. On the other hand, if subjects are involved in a discrimination task requiring nonphonetic (e.g. acoustic) analysis of speech or 20 nonspeech stimuli, a hemispheric asymmetry should not be obtained, even if the evoking stimulus is verbal in nature. Thus, by averaging potentials to a speech and a nonspeech stimulus within the same series of stimuli, it can be determined if the data reported by Wood (Wood, 1975; Wood et al., 1971) were due to a differential left hemisphere activation in preparation for phonetic analysis or to the actual processing itself. Finally, since speech and nonspeech stimuli will be employed and subjects will be discriminating stimuli belonging to one class (i.e. speech or nonspeech) from other stimuli of both classes, these experiments will allow for investigating possible differences among potentials when evoked by a relevant stimulus, an irrelevant stimulus belonging to the same class as the relevant stimulus, and an irrelevant stimulus belonging to the irrelevant class. Ford, Roth, Dirks and Kopell (1973) reported that an irrelevant stimulus in the same sensory modality as a relevant one results in a smaller amplitude P3 than does the relevant stimulus, but that no difference in N2 occurs between the two stimuli. In contrast, waveforms evoked by stimuli in a modality other than the one of the relevant stimulus exhibit a small, if any, N? and P_. Because of these findings, those authors suggested that the N component reflects preliminary processing of stimuli based on sensory modality and that the P component reflects the final decision process. Interpreting their data in terms of a neural template model (Hillyard, Squires, Bauer, & Lindsay, 1971) , they suggested that an irrelevant 21 stimulus of the same modality caused a P wave because of a partial match to the template of the relevant stimulus. The intramodality auditory stimuli employed by Ford and her colleagues were not clearly described in their article. They used a click as one stimulus and a change in the level of a background noise as the other stimulus, but they did not report the duration of the click or the duration of the new noise level before it returned (if it did return) to the original "prestimulus" level. Also, they did not report the transition time from one noise level to the other. These stimulus parameters are important to the perception of the stimuli and without them it is difficult to evaluate the degree of "match" of one stimulus to a neural template of the other. The present work involves procedures which use two speech stimuli and two nonspeech stimuli. Since these four auditory stimuli will be presented to subjects while they are engaged in tasks of stimulus relevance, the present study should serve as a more thorough test of the template mode. Since irrelevant stimuli will occur in both a relevant and an irrelevant class, a template model would predict an ordering of the P3 amplitudes as follows: the relevant stimulus should give the largest P,, followed by the irrelevant stimulus in the same class as the relevant stimulus, which, in turn, should give a larger P3 than an irrelevant stimulus in an irrelevant class. 2 2 If the N component reflects a decision of sensory modality as suggested by Ford, then no difference in the N amplitude should occur in the various conditions of stimulus relevance because all stimuli are auditory. Summary There is a substantial body of research data, gathered from both clinical and nonclinical behavioral investigations, which support the phenomenon of specialization of function in the human brain. In addition, recent findings suggest that there are structural asymmetries in the cortex which parallel the functional differences. Results of recent electrophysiological investigations suggest that the left hemisphere's specialization for langugage is reflected in the auditory evoked potentials elicited by speech and nonspeech stimuli, and during linguistic and nonlinguistic processing of speech stimuli. Because of several inconsistencies in the data, however, the precise nature of the effect as well as the actual processing mechanisms reflected in the asymmetries remain obscure. One purpose of the present work is to further investigate the evoked potential correlates of hemispheric specialization for language by recording auditory evoked potentials to speech and nonspeech stimuli from left and right temporo-parietal recording sites. The results of the present investigation should clarify certain issues regarding hemispheric asymmetries, with the following questions being addressed: How do left hemisphere responses compare to right hemisphere 23 responses when the evoking stimulus is a speech signal and when it is a nonspeech signal? Are individual subjects' relative left-right asymmetries influenced by the verbal nature of a stimulus? Do subjects' responses to stimuli differ if they are evoked during a phonetic processing task as compared to a non- phonetic processing task? Finally, can the reported hemispheric asymmetries be explained as a preparatory activation of the left cortex due to the anticipation of phonetic processing of stimuli? A second issue to which the present work is addressed is one of electrophysiological correlates of stimulus relevance, and the results should broaden our understanding of the N-> and P components * 3 of the evoked potential with respect to their psychological corre- lates. For example, how do they relate to stimulus modality or stimulus similarity? Data from the present work should support or refute the interpretations that N2 reflects a decision regarding sensory modality of a stimulus and that P3 reflects a higher level match-mismatch detector. METHOD Subjects Subjects were six males and two females who were paid $2.00 per hour for their participation in the study. Six subjects were recruited through an ad in the University of Florida student news- paper and two subjects were acquaintances of the investigator. Subjects' ages ranged from 19 years to 27 years, with a mean of approximately 23 years. All subjects had normal hearing and no history of hearing problems. All subjects were self-reported right- handers, with six of the eight having two parents who were also right-handed. The remaining two subjects had one right-handed and one left-handed parent. Stimuli and Apparatus Stimuli consisted of three natural speech consonant-vowel syllables and two nonspeech signals. The speech stimuli consisted of a low-pitched /ba/ having a fundamental frequency of 110 Hz; a low-pitched /da/ also with a fundamental frequency of 110 Hz; and a high-pitched /ba/ with a fundamental frequency of 144 Hz. Each stimulus was uttered by an adult male and was 260 msec in duration (+ 4 msec). Fundamental frequency was measured from displays of the waveforms on a storage oscilloscope. 24 25 The nonspeech stimuli were two square wave signals generated by a General Radio oscillator, Type 1313-A. One signal had a frequency of 110 Hz and the other a frequency of 144 Hz. The nonspeech signals were shaped with an electronic switch (built at the Communication Sciences Laboratory of the University of Florida) to have a rise-fall time of 10 msec and a duration of 260 msec. Each stimulus was recorded onto a Sony two-channel stereo tape recorder, Model TC-353D and then dubbed onto an identical tape recorder in a prescribed order to make four series of stimuli, each on a separate tape. Each of the four audiotapes presented a total of 256 stimuli: 64 presentations each of two speech stimuli and two nonspeech stimuli. The order of the four types of stimuli on each tape was randomized with two restrictions: first, each quarter of the tape (64 stimuli) contained an equal number (16) of the four stimuli; second, no more than two consecutive presentations of a given stimulus were allowed. These restrictions served to eliminate extreme sequences of stimuli which might have occurred by chance. The stimuli were recorded to have an interstimulus interval ranging from 3 sees to 6 sees. The orderings of the stimuli on the tapes are shown in Appendix A. Two of the audio tapes contained the following four stimuli: the two nonspeech stimuli described above and the two low-pitched speech stimuli, /ha/ and /da/. These two tapes are referred to as 26 phonetic tapes. Two phonetic tapes having different orderings of stimuli were employed to reduce the possibility of the subjects becoming familiar with the order of the stimuli. The remaining two audio tapes contained the same two nonverbal stimuli, but the speech stimuli were the low-pitched /ba/ and the high-pitched /ba/. These tapes are referred to as the nonphonetic tapes. Again, two tapes were prepared to reduce subjects' familiari- zation with the order of the stimuli. The low frequency and high frequency nonspeech stimuli are referred to throughout this paper as T. and T , respectively. The low /ba/ is referred to as CV, . CV 2 refers to the remaining consonant-vowel stimuli, i.e., the low /da/ on the phonetic tapes and the high /ba/ on the nonphonetic tapes. Responses were averaged, using Digital's Basic Averager, to two evoking stimuli — CV. and T . All stimuli were played on a Sony stereo tape recorder, Model TC-353D and presented to the subjects binaurally at 35 dB SL through Grason-Stadler TDH-39 earphones. The evoked potentials were recorded using Grass silver-silver chloride electrodes and Grass electrode paste. The signals were amplified by calibrated Grass P511 amplifiers with a gain of 20,000 and half-amplitudes of .1 Hz and 100 Hz. Responses, stimuli, and synchronized pulses were all recorded onto separate channels of an Ampex FM tape recorder, Model Fr-IOOA. Averaging to CV was on line; averaging to Tl was off line, without further filtering. Averages were plotted on a Houston Omnigraphic X-Y plotter. 27 Procedure Subjects were tested individually over three separate sessions. They were seated comfortably in a reclining chair in a sound-proof IAC room. Recordings were taken from the vertex (Cz) and the left and right temporo-parietal scalp locations (midway between T3 and P3 on the left and T4 and P4 on the right) after carefully prepar- ing the site and applying the electrodes. The active sites were prepared by removing the hair at the location, rubbing briskly with acetone, rubbing the area with a small amount of paste, apply- ing the electrodes and securing them with collodion. The removal of a few strands of hair at each site not only lowered the resist- ance but also guaranteed precise repositioning of the electrodes in subsequent sessions. The scalp locations were referenced to linked mastoids which were prepared similarly but with the elec- trodes secured with tape. The resistance between each pair of electrodes was kept below 5000 ohms and approximately equal (within 1000 ohms) for each pair. The subject was grounded with a mid- forehead electrode. The same amplifier was used for the vertex recordings for all subjects. However, to control for any slight differences between the amplifiers used for the hemispheric recordings the leads from half the subjects were reversed so that each of the two remaining amplifiers received signals from the left hemisphere for half the subjects and the right hemisphere for the other half of the subjects. 28 Each subject participated in eight conditions of stimulus relevance (four with the phonetic tapes and four with the non- phonetic tapes) and two control conditions (one with each type of tape). The ten conditions are summarized in Table 1. Prior to each condition of stimulus relevance, subjects were instructed as to the identity of the relevant stimulus for that trial. Subjects were told that they should listen selectively for the relevant stimulus and ignore all other stimuli. Half of the subjects responded, following the presentation of a relevant stimulus, by depressing with their index fingers a small button mounted on a box on the arm of their chair. The remaining half of the subjects responded with a similar button-push following all irrelevant stimuli and withheld a response following each presen- tation of the relevant stimulus. To avoid the contamination of the evoked response by muscle activity or motor potentials, subjects were instructed to wait from 1 to 2 sees following the stimulus before responding. In order to equalize the responses between the left and right hands, midway through each trial (following presen- tation of 128 stimuli) the experimenter switched the response box from one arm of the subject's chair to the other. Also during midtrial, the experimenter reversed the subject's headphones so that the effect of any possible difference between the earphones would be equalized. In communicating to the subjects, the experimenter identified the relevant stimulus for a given trial as the "low tone," "high 29 Table 1 Summary of Experimental Conditions Task Relevant Stimulus Phonetic Nonphonetic Evoking Stimulus: CV CV, CV, Control Evoking Stimulus: T, 2 cv1 CV 2 Control Note: Dependent variables in all conditions were amplitude and latency measures for N, , P_, N?, and P^ components of the evoked potential recorded from vertex, left temporo-parietal , and right temporo-parietal sites. In every condition the subject was presented with a series of 256 stimuli (64 each of CVi , CV^ and T ) and potentials were recorded to CV, and T . 1' 30 tone," "low /ba/," "low /da/" (on phonetic tapes), or "high /ba/" (on nonphonetic tapes) . During the control conditions subjects were instructed to read and not attend to any of the stimuli. No response was required of the subjects during the control conditions. Half the subjects were presented first with the phonetic tapes and second with the nonphonetic tapes. This order was reversed for the other subjects. To minimize the effects of fatigue or habituation on the evoked potentials, the temporal order of the stimulus relevance conditions was controlled. For a given subject the temporal position of the "CV, relevant" condition, the "T, relevant" condition, and the control condition were identical for phonetic and nonphonetic tape presentations and the CV2 relevant" and "T relevant" conditions were never more than one position removed for the two types of tapes. The actual order ings of conditions for the subjects are shown in Appendix B. Each experimental session lasted from two to three hours: approximately 30 minutes for application of electrodes and about 90 minutes to 2 hours of actual recording. Subjects were given a 15 to 20 minute break following the first 60 minutes of recording. Prior to each recording session subjects were instructed to remain very still, not to swallow, and not to blink during the recording sessions. If it was absolutely necessary for them to do one of these things they were to do it at the same time they pushed 31 the button (between the presentation of stimuli) . During the conditions of stimulus relevance subjects fixated on a picture or an object so that eye potentials would not contaminate the evoked responses. RESULTS Auditory evoked potentials to a speech stimulus (CV ) and a nonspeech stimulus (T^) were recorded from vertex, left temporo- parietal, and right temporo-parietal locations. The data were analyzed for topographical differences, effects of phonetic process- ing, and effects of stimulus relevance. Peak amplitudes of the Nl' P2' N2' and P3 comPonents' tne Ni-P2 Peak-to-peak amplitude, and the latencies of the various components were measured. The 6 msec following stimulus onset was used as the baseline from which the peak amplitudes were measured. Components were defined as follows: N, was the most negative point occurring between 80 and 140 msec following stimulus onset; P? was the most positive point occurring between 170 and 240 msec following stimulus onset; P was the most positive point between 260 and 420 msec post- stimulus onset; and N~ was the most negative point occurring between P? and P-, . To determine the significance of the differences among mean amplitudes or latencies of the components recorded from the three locations during phonetic and nonphonetic processing tasks, repeated measures analyses of variance were computed for various condition of stimulus relevance. Because of the design of the study, for each evoking stimulus there is one condition in which the evoking stimulus 32 33 is the relevant stimulus (referred to as R/R) ; one condition in which the evoking stimulus is an irrelevant stimulus but of the same class (i.e. speech or nonspeech) as the relevant stimulus (I/R) ; and two conditions in which the evoking stimulus is irrele- vant and of the irrelevant class (I/I) . The two I/I conditions for the CVj evoking stimulus are the Tj relevant and the T2 rele- vant conditions; for the T, evoking stimulus the two I/I conditions are the CV-^ relevant and CV_ relevant condtions. In order to simplify the analyses and eliminate redundant information, the second I/I condition of stimulus relevance was excluded. This leaves a total of four conditions: R/R, I/R, I/I, and Control. For the CV-, evoking stimulus these are the CV, relevant, CV2 relevant, T relevant, and Control; for the T, evoking stimulus these are the T, relevant, T2 relevant, CV-, relevant, and Control conditions. When appropriate, the Sign test (Siegel, 1956) was used in addition, to determine the significance of the number of persons showing a particular response characteristic. The findings of these analyses are reported below. Topographical Effects Amplitude measures. Peak amplitudes of the N-, and P2 waves, as well as the peak-to-peak N^-P2 amplitudes were substantially larger at the vertex than they were at the lateral sites. This effect is illustrated in Figures 1 and 2, which show the evoked 34 Relevant Stimulus CV, cv. 100 msecs Figure 1. Averaged evoked potentials to CVj_ stimulus recorded from vertex (V) , left (L) , and right (R) temporo-parietal sites during various conditions. Each potential is a com- posite of eight subjects. Calibration = 10 microvolts. 35 Relevant Stimulus cv\ Control 100 msecs Figure 2. Averaged evoked potentials to T-^ stimulus recorded from vertex (V), left (L) , and right (R) , temporo-parietal sites during various conditions. Each potential is a com- posite of eight subjects. Calibration = 10 microvolts. 36 potentials recorded from vertex, left temporo-parietal and right temporo-parietal locations during each condition of stimulus relevance. Figures 1 and 2 show the potentials evoked hy CV-^ and T , respectively. Each waveform in the figures represents a composite of responses from all subjects in both phonetic and non- phonetic conditions. The actual values of the amplitudes of the various components for each location are presented in Table 2. The differences between the vertex responses and the temporo- parietal responses are significant at the .01 level of probability for every condition of stimulus relevance as well as for the control conditions. Results of Tukey's a posteriori test (Kirk, 1968) com- paring the mean amplitudes of the N,, P , or N -P2 components recorded from the left location with those of the riqht location were all nonsignificant. No significant topographical differences were obtained among the three recording sites for the N2 or P3 com- ponents in any condition. Although the overall mean differences between the N , P , and N,-P amplitudes between the left and right hemispheres did not differ significantly, in most conditions the majority of persons did show asymmetries with larger amplitudes on the left. Table 3 presents the number of persons in each condition whose left hemisphere amplitudes were larger than the right hemisphere ampli- tudes. Although the majority of persons did show a larger left hemisphere response, particularly when CV-, was the relevant 37 Table 2 Mean Amplitudes (microvolts) for Major Components of Evoked Responses Recorded from Three Locations Evoking Stimulus CV. Com- Relevant ponent Stimulus Vc L R Relevant Stimulus V I, R T 1 12 3 9.7 9.5 T2 13 4 10.4 10.1 cv1 13 2 10.3 9.5 Control 12 1 10.0 8.9 CV, CV, Control 10.5 8.5 8.1 9.4 8.3 7.4 9.6 7.7 7.7 9.6 8.2 7.9 CV, CV, Control 10.1 6.9 6.5 9.1 6.2 6.3 10.5 6.7 6.6 6.2 4.1 4.3 CV, Control 13.1 8.3 7.7 12.5 8.7 7.7 11.9 7.6 7.3 9.0 5.7 5.7 N1"P2 CVn CV- Control 20.6 15.4 14.6 18.5 14.5 13.7 20.1 14.4 14.3 15.8 12.3 12.2 CV, Control 25.4 18.0 17.2 25.9 19.1 17.8 25.1 17.9 16.8 21.1 15.7 14.6 Table 2 - continued Evoking Stimulus cv. Com- ponent Relevant Stimulus V3 L R Relevant Stimulus V L R N2 CV1 .7 .5 .8 Tl .7 -.1 -1.5 cv2 -.2 -.2 .1 T2 -1.9 -1.1 -1.0 Tl .9 .7 .7 CV1 -1.1 -1.0 -.8 P3 cv1 4.2 4.5 4 .9 Tl 4.8 3.4 4.4 cv2 2.1 2.3 3 .2 T2 3.3 2.4 2.5 Tl 3.5 3.6 3 9 CV1 1.4 2.2 3.4 Note: A positive amplitude for N2 means that the most negative peak between P0 and P., was above the baseline. 2 """ "3 a,,_ V=Vertex; L=Left Temporo-Parietal ; R=Right Temporo-Parietal. 39 Table 3 Number of Persons Having Greater Left Hemisphere Amplitudes for Various Components Component Relevant Stimulus N1~P2 Evoking Stimulus: CV^ cv, cv^ Control Evoking Stimulus: T^ CV, Control 40 stimulus, in only one condition did the Sign test on this frequency- data actually reach significance. This condition is the one in which T.-, was relevant and the evoking stimulus was T . In this *■ 1 condition a significant number of persons (seven out of eight) had a larger P2 amplitude on the left than on the right. To provide an indication of the degree of asymmetry which occurred for each subject, the procedure described by Matsumiya et al. (1972) and later employed by others (Friedman et al . , 1975a; Molfese et al., 1975) was used. This procedure entails calculating R-values for each subject by dividing the left hemisphere amplitude by the sum of the amplitudes of the left and right hemispheres. Thus, an R- value greater than .5 indicates a larger left hemisphere N1~P2 resP°nse while a value less than .5 indicates a larger right hemisphere response. R-values for each subject are shown in the upper half of Figure 3. Each R-value shown is a composite of all conditions. Since R-values are simply ratios, however, they do not provide a clear indication of the degree of asymmetry. For example, a 2 microvolt difference between the left and right amplitudes would yield substantially larger R-values if the amplitudes were around 5 microvolts than if they were around 20 microvolts. Also, a person whose left hemisphere amplitude is 3 microvolts larger than the right but whose amplitudes are 20 and 17 microvolts for the left and right hemispheres, respectively, would have an R-value of .55 while a second person having only a 1 microvolt difference 41 8 x SUBJECT SUBJECT Figure 3. R-values (upper) and hemispheric voltage differences (lower) for all subjects across all conditions. A larger left hemisphere amplitude is represented by an R-value > . 5 or a posi- tive voltage difference. 42 would have the same ratio if his left and right amplitudes were about 7 and 6 microvolts, respectively. Because of this fact, it was desirable to calculate actual magnitudes of the hemisphere differences as indices of asymmetries in addition to the R-values. Thus, shown in the lower half of Figure 3 are the differences between the mean left hemisphere N -P_ amplitude across all conditions and the mean right hemisphere N.-P- amplitude across all conditions. It is comforting to see that the two halves of the figure agree fairly well with respect to individual subjects' asymmetries, but it is the lower half which provides information regarding the abso- lute magnitude of each person's asymmetry. Latency measures. In general, the latencies for the N and P_ components of the responses recorded from the vertex were slightly shorter than those recorded from the lateral sites. Across all conditions the mean N latency for vertex, left, and right temporo- parietal locations are 107, 108, and 109 msec, respectively, when CV was the evoking stimulus; and 106, 109, and 110 msec when T was the evoking stimulus. The corresponding latencies for P~ are 200, 204, and 205 msec for CV ; and 190, 194, and 195 msec for T-,. Latencies for individual conditions of stimulus relevance are presented in Table 4. The analyses of variance failed to show any significant differences in the mean latencies for the three recording sites, but it is notable that in no condition were the mean N or P^ latencies for either of the temporo-parietal sites less than that for the vertex. 43 Table 4 Mean Latencies (msec) for Major Components of Evoked Responses Recorded from Three Locations Evoking Stimulus CV1 Com- Relevant Relevant ponent Stimulus Va L R Stimulus V L R Nj CV1 107 107 109 T. 106 108 107 CV2 108 110 110 T2 107 109 110 Tx 106 107 108 CVX 106 109 110 Control 106 107 107 Control 106 108 111 P2 CV1 201 203 204 Tl 189 192 190 cv2 199 201 204 T2 188 193 196 Ti 201 205 208 CVi 190 198 196 Control 200 208 204 Control 192 194 197 N2 CV1 278 274 272 Tl 276 278 279 cv2 290 287 282 T2 285 279 282 Tl 292 288 283 cv2 284 277 276 P3 cvl 325 335 338 Tl 328 336 335 ^2 320 323 326 T2 334 334 339 Tl 326 327 326 cv1 324 329 329 av= Vertex; L=Left Temporo-Parietal ; R=Right Temporo-Parietal . 44 The mean latencies for the P, component of the evoked potentials were also shorter for the vertex than for the left or right locations, but these differences also failed to reach statistical significance. In contrast to the other components, the latencies of the N2 wave were generally longer when recorded from the vertex. The latency values for N and P are shown also in Table 4. No consistent latency differences occurred between the left and right recording sites for any of the components. Phonetic versus Nonphonetic Tape Effects Amplitude measures. The mean amplitude of the various compo- nents of potentials evoked by CV, and T. on the phonetic tapes and the nonphonetic tapes are shown in Table 5 for each condition of stimulus relevance. The values in that table represent the mean amplitudes for the three recording sites across all eight subjects. The only differences that proved to be significant occurred during the condition in which subjects were discriminating the CV stimulus from the other three stimuli. In that condition, the amplitude of the N component was significantly larger during phonetic processing of the speech stimuli (10.3 microvolts) than during non- phonetic processing of them (7.8 microvolts). This difference is significant at the .05 level of probability, F (1,7) = 10.90. As described under the procedure section of this paper, the evoked response to the nonspeech stimulus, T-, , was averaged during the same series of stimulus presentations as was the response to 45 Table 5 Mean Amplitudes (microvolts) for Major Components of Potentials Evoked by Stimuli on Phonetic and Nonphonetic Tapes Evoking Stimulus Com- Relevant CV1 Relevant Tl ponent Stimulus Pa N Stimulus P N Nl CV1 10.3 7.8 Tl 10.5 10.5 ^2 8.1 8.7 T2 12.0 10.5 Tl 8.3 8.5 CV1 11.0 10.9 Control 9.7 7.3 Control 10.6 10.1 P2 CV1 6.6 9.0 Tl 10.1 9.6 cv2 7.4 7.0 T2 9.1 10.2 Tl 8.2 7.7 cv1 9.1 8.8 Control 3.9 5.8 Control 6.2 7.4 Nrp2 CV1 16.9 16.8 T 1 20.6 20.1 cv2 15.5 15.7 T2 21.1 20.7 Tl 16.5 16.2 CV1 20.1 19.7 Control 13.6 13.1 Control 16.8 17.5 46 Table 5 - continued Evoking Stimulus Relevant Stimulus cv1 Relevant Stimulus Tl Com- ponent pa N P N N2 CV1 0.0 1.3 Tl 1.0 -1.6 CV2 0.3 -0.5 T2 -2.3 -0.4 Tl 1.4 0.1 CV 1 -0.8 -1.2 P3 CV1 4.0 5.0 T 1 4.9 3.8 cv2 2.4 2.6 T2 2.7 2.8 T 1 4.5 2.8 CV1 2.8 1.9 Note: A positive amplitude for N2 means that the most negative peak between P2 and P was above the baseline. aP=Phonetic Tape; N=Nonphonetic Tape. 47 CV.. ; yet there was virtually no difference between the N amplitudes of potentials evoked by T, on the phonetic tapes and on the nonphonetic tapes. The mean amplitudes are 11.0 and 10.9 microvolts, respectively, F_ (1, 7) = .02, p>.05. In the same condition, a similar effect occurred with the amplitude measure of the P2 component, but in the opposite direc- tion. During phonetic processing the mean peak amplitude of the response to CV was 6. 6 microvolts and during nonphonetic proces-' sing it was 9.0 microvolts, F (1, 7) = 10.51, p-d05. The cor- responding amplitudes of the potentials evoked by T, are 9.1 and 8.8 microvolts, F (1, 7) = .15, p>.05. None of the other components of the potentials were found to be significantly different when evoked by stimuli on the phonetic tapes as compared to when evoked by stimuli on the nonphonetic tapes. Latency measures. Latencies of components elicited by stimuli on the two types of tapes are shown in Table 6. The only signifi- cant difference between the phonetic and nonphonetic processing tasks was in N latency and occurred for the CV evoking stimulus 1 1 when CV-^ was the relevant stimulus. Across all subjects, the mean latencies for N are 110 and 105 msecs for the two tasks, respectively, F_ (1, 7) = 6.23, p<.05. In contrast, the corresponding latencies for responses evoked by T were not significantly different from each other, F (1, 7) = .44, p>.05. These values are 109 msecs and 108 msecs for phonetic and nonphonetic tapes, respectively. 48 Table 6 Mean Latencies (msec) for Major Components of Potentials Evoked by Stimuli on Phonetic and Nonphonetic Tapes Evoking Stimulus CV, Com- Relevant Relevant ponent Stimulus Pa N Stimulus P N Nl CV1 110 105 Tl 108 106 CV 2 105 113 T 2 107 110 Tl 105 109 CV1 109 108 Control 106 107 Control 110 107 P2 cv1 203 202 Tl 192 188 cv2 196 207 T2 190 194 Tl 205 205 CV1 195 194 Control 206 202 Control 196 193 N 2 cv1 274 276 Tl 276 274 cv2 288 285 T2 281 284 Tl 285 290 CV1 280 278 P3 cvl 334 332 Tl 327 339 cv2 319 328 T2 335 337 T 1 323 330 cv1 328 327 P=Phonetic Tape; N=Nonphonetic Tape. 49 A significant interaction occurred between processing require- ments and recording locations during the condition in which CV\ was relevant and CV, was the evoking stimulus, F (2,14) = 5.19, p<05. This interaction is illustrated in Figure 4. It can be seen from that figure that the latency difference between phonetic and nonphonetic processing was greatest when recorded from the left hemisphere, moderate from the vertex, and least from the right hemi- sphere. No signficant latency differences were obtained for the other evoked potential components in that condition or for any of the components in other conditions of stimulus relevance. Other measures. The procedure described by Wood et al. (1971) and subsequently by Wood (1975) and by Galambos et al . (1975), was employed in the present study to test for differences in the general shape of the waveforms obtained during phonetic and nonphonetic processing of stimuli. For each hemisphere, a Wilcoxen matched- pairs signed-ranks test (Siegel, 1956) was performed on the digi- tized output voltages at each of 180 time points over a duration of 540 msecs. Each time point represents an interval of 3 msecs. To reduce the probability of a Type I error, a .02 probability level was chosen for the Wilcoxen tests. Since 180 tests were performed for each pair of averages, it would be expected that approximately four of the tests for each pair would reach signifi- cance on the basis of chance. 50 TASK PHONETIC 0| NONPHONETIC fe^j 112 VERTEX RIGHT RECORDING SITE Figure 4. Latencies for N-j, component of potentials evoked by CV]^ during phonetic and nonphonetic processing tasks. 51 The pairs of averages on which the Wilcoxen tests were per- formed are shown in Figure 5. The upper tracings in each of the four sections of that figure are averages of potentials recorded from the left hemisphere; the lower tracings are from the right hemisphere. The solid lines are averages of potentials evoked by stimuli on the phonetic tapes; the dotted lines by stimuli on the nonphonetic tapes. Each pair of averages is anchored at the 6 msec baseline. Figure 5a and 5b show averages recorded during the conditions in which CV was the relevant stimulus. In Figure 5a are potentials evoked by CV ; in Figure 5b are potentials evoked by T.. Figure 5c and 5d show corresponding averages obtained during the conditions when T was the relevant stimulus. It can be seen in Figure 5a that, while the right hemisphere's responses to CV, are virtually identical when evoked by stimuli on phonetic and nonphonetic tapes, the left hemisphere's responses to those stimuli are relatively dissimilar. The absolute difference obtained is not a large one, only about 2 microvolts. But considering that the mean peak-to-peak amplitudes are about 13 microvolts, this represents a 15 percent difference. This difference suggests that the left hemisphere's activity in response to a speech sound differs when the stimulus is processed phonetically. There is little differ- ence in either hemisphere's response to T in that condition. Table 7 shows the number of time points at which differences between the potentials evoked by phonetic and nonphonetic tapes 52 RELEVANT STIMULUS cv. b. L CV cv. C. L d. L phonetic nonphonetic 100 msecs Figure 5. Averaged evoked potentials recorded from left and right hemispheres during phonetic and nonphonetic tasks. Calibration = 10 microvolts. 53 Table 7 Significant Wilcoxen Tests between Potentials Evoked by Stimuli on Phonetic and Nonphonetic Tapes Hemisphere Evoking Stimulus Left Right Relevant Stimulus: CV CV. 35 Tl CV, Tl Relevant Stimulus: T 54 were found to be significant. It can be seen that the left hemi- sphere's response to the CV stimulus on the two tapes differed significantly at several points along the time continuum during the condition when CV, was relevant. This is particularly true in the region of the P2 component — 18 significant differences occurred between 170 msecs and 240 msecs following stimulus onset. Not only is 35 significant points out of 180 substantially above the chance number of four, but because the significant points tend to be clustered in certain regions of the waveform and not distri- buted randomly buttresses the validity of the difference. In contrast, the number of points which were significantly different for the right hemisphere's averages in the CV^ relevant condition for the CV^ evoking stimulus is six, very close to chance expectation. There were no significantly different points for either the left or the right hemisphere responses evoked by Ti . Comparable tests were computed when Tj_ was the relevant stimulus. The results of these tests are shown in Table 6 also. It can be seen that neither hemisphere responded differentially to the CV-. stimulus when T^ was relevant. In other words, when T]_ was the stimulus to be discriminated it made no difference whether the speech stimuli on the two tapes were phonetically dif- ferent or only acoustically different. Again, few differences occurred between the averages evoked by T,. Figure 5c and 5d illustrate these findings. 55 Effects of Stimulus Relevance Amplitude measures. Amplitudes of vertex potentials were analyzed for the general effects of stimulus relevance after com- bining the data for phonetic and nonphonetic conditions. Amplitudes of the components are presented in Table 8. Significant main effects due to stimulus relevance were obtained in the analyses of the N^HP- peak-to-peak amplitudes and of the P,, N2 and P3 peak amplitudes. The values are shown in Table 8. In that table the column labelled R/R shows data obtained when the evoking stimulus, either CV or T1 , was the relevant stimulus; I/R shows data obtained when the evoking stimulus was an irrelevant stimulus but of the same class (i.e. speech or nonspeech) as the relevant stimulus; I/I shows data when the evoking stimulus was irrelevant and of the irrelevant class. Results of statistical tests for the various components of the potentials are summarized below. 1- N -P components. Results of Tukey's a posteriori tests 1 * showed that the mean amplitude of the N1~P2 component was signifi- cantly smaller in the control conditions than in all of the conditions of stimulus relevance, p<. 01. The three stimulus relevance condi- tions did not differ significantly from one another. Since no difference among conditions was obtained for N, , the differences in N1_p2 are due to changes in P in the various conditions. 2. p component. The P component of the control conditions was significantly smaller than in the other conditions, p<-01, but 56 Table 8 Mean Amplitudes (microvolts) of Evoked Potential Components in Condi- tions of Stimulus Relevance Condition Com- ponent R/Ra I/Rb I/Ic Control F(3, 21) P Nl 11.4 11.5 11.4 10.8 .49 n. s. P2 11.6 10.8 11.1 7.6 9.10