Tone discrimination performance in schizophrenic patients and normal volunteers: impact of stimulus presentation levels and frequency differences

PSYCHIATRY RESEARCH Psychiatry Research 57 (1995) 75-82

Tone discrimination performance in schizophrenic patients and normal volunteers: impact of stimulus presentation levels and frequency differences Henry H. Holcomb*“, Eva K. Ritzl”, Deborah R. Medoff”, Jonathan Nevitt”, Barry Gordon b, Carol A. Tammingaa “Department of Psychiatry, University of Maryland School of Medicine, Maryland Psychiatric Research Center, P.O. Box 21247. Baltimore, MD 21228. USA bJohns Hopkins Medical Institutes. Baltimore, MD 21205, USA

Received 12 July 1994; revision received 16 August

1994; accepted

7 November 1994

Abstract Psychophysical and cognitive studies carried out in schizophrenic patients show high within-group performance variance and sizable differences between patients and normal volunteers. Experimental manipulation of a target’s signal-tonoise characteristics can, however, make a given task more or less difficult for a given subject. Such signal-to-noise manipulations can substantially reduce performance differences between individuals. Frequency and presentation level (volume) changes of an auditory tone can make a sound more or less difficult to recognize. This study determined how the discrimination accuracy of medicated schizophrenic patients and normal volunteers changed when the frequency difference between two tones (high frequency vs. low frequency) and the presentation levels of tones were systematically degraded. The investigators hypothesized that each group would become impaired in its discrimination accuracy when tone signals were degraded by making the frequencies more similar and the presentation levels lower. Schizophrenic patients were slower and less accurate than normal volunteers on tests using four tone levels and two frequency differences; the schizophrenic patient group showed a significant decrement in accuracy when the signal-to-noise characteristics of the target tones were degraded. The benefits of controlling stimulus discrimination difficulty in functional imaging paradigms are discussed. Keywords:

Auditory

discrimination;

Attention;

Psychophysical

1. Introduction

Psychophysical and cognitive studies with schizophrenic patients (Mirsky and Kornetsky, 1968; Spohn et al., 1977; Kugler and Caudrey, * Corresponding author, Tel: +I

410 455-7915; Fax: +1 410

788-3394.

0165-1781/95/.$09.50 0 1995 Elsevier Science Ireland SSDI 0165-1781(95)02270-D

parameters

1983; Cohen et al., 1987, 1988; Goldberg et al., 1989; Nestor et al., 1991) show high within-group performance variance and sizable differences between patients and normal volunteers. Cognitive challenge conditions are often used in conjunction with dynamic brain imaging to identify taskrelated brain areas (Cohen et al., 1987, 1988). Wide performance variations between individuals

Ltd. All rights

reserved

76

H.H. Hoicomb et al. /Psychiatry Research 57 (1995) 75-82

performing the same task make it difficult to interpret physiological measures obtained from dynamic imaging studies (Weinberger et al., 1986). By reducing between-subject variability in task performance, it is easier to interpret the physiological changes associated with the specific behaviors. Animal and human studies suggest that task difficulty and relevance substantially contribute to the neural activity patterns in cortical association regions (Hyvarinen et al., 1980; Moran and Desimone, 1985; Richmond and Sato, 1987; Spitzer et al., 1988; Meyer et al., 1991; Spitzer and Richmond, 1991). Stimulus relevance and discrimination difficulty will vary between subjects. These differences contribute to physiological variability between groups and individuals, and complicate physiological comparisons between normal volunteers and schizophrenic patients. Intersubject performance differences can be reduced if test stimulus characteristics are modified in conjunction with the subject’s responses. It is our expectation that physiological variability can be reduced by minimizing the widely differing levels of stimulus discrimination difficulty between subjects. This report describes our initial efforts to modify sound stimuli in a way compatible with this goal. Many task-related stimuli can be modified along a continuum (frequency, loudness, duration, and delay). Degradation or enhancement of stimulus features makes it possible to study subjects performing the same task at different signal-to-noise (S:N) ratios (Nuechterlein et al., 1983; Nestor et al., 1991). By systematically varying the S:N ratios used in a series of tasks, one can obtain a series of measurements of the subject’s responses to increasing levels of task difficulty. Although subjects may have different absolute levels of detection accuracy when responding to identical S:N calibrated stimuli, they may exhibit similar responses to changes in the S:N ratio (Tanner, 1964). Performance measurements made in association with systematic S:N changes provide the necessary information to compute and compare the discrimination accuracy of diverse groups. The investigator can then determine to what extent the groups differ in their responses to changes in the S:N ratio.

The forced-choice auditory discrimination study presented here was designed to assess whether schizophrenic subjects and normal volunteers would show similar performance response trends in association with identical S:N changes. We predicted that although globally the schizophrenic group would perform more slowly than the normal volunteers, the two groups would exhibit similar response trends when we tested them with incremental shifts in the S:N ratio. Cardozo’s (1974) auditory discrimination curves in normal subjects reveal striking linear trends with regard to signal detection over a range of S:N levels. By making a target tone softer in presentation level (reducing volume) or by increasing the background noise level, Cardozo made trained voluntary observers progressively more impaired in their efforts to discriminate between one frequency and another. Studies by Tanner (1964) showed that when the frequency separation between two tones dropped below 100 Hz, a subject’s ability to recognize the target tone as relatively high or low frequency became faulty in a predictable linear trend. Our study extends the psychophysical discrimination work performed in normal subjects to medicated schizophrenic patients. We report data that point up the differences between schizophrenic patients and normal volunteers in of a forced-choice tone the performance discrimination task over a range of four presentation volume levels and two frequency differences. 2. Methods 2.1. Demographic information Nine normal volunteers had a mean age of 34.4 years (SD = 6.5); six were men and three were women. All were free of psychiatric and medical illness; none had first degree relatives with a diagnosable mental illness. They were taking no medications with central nervous system activity. Ten male and one female schizophrenic patients were housed on research units at the University of Maryland Psychiatric Research Center. All patients were optimally medicated. Patients had a mean age of 29.5 (SD = 9.6) and a mean Brief Psychiatric Rating Scale (BPRS; Overall and

H.H. Holcomb et al. /Psychiatry Research 57 (1995) 75-82

71

Table I Patient demographics Patient No.

Gender

Age (years)

Medication

BPRS total score

8 9 10 I1

Male Male Male Female Male Male Male Male Male Male Male

28 38 31 33 31 38 29 26 26 I9 40

Haloperidol + benztropine Thioridazine + benztropine Haloperidol + benztropine Clozapine Clozapine Haloperidol + benztropine + diazepam Haloperidol + benztropine Carbamazepine + stelazine + benztropine Haloperidol + thyroid + benztropine Haloperidol Amitriptyline + perphenazine

35 36 47 35 31 25 33 46 58 38 35

Note. BPRS, Brief Psychiatric Rating Scale.

Gorham, 1962) score of 38.1 (SD = 8.6) (see Table 1). 2.2. Testing design Experiment environment. Subjects were studied in a quiet, isolated room (8 x 12 feet). Testing occurred between 10:00 h and 14%) h; medications were given each morning at 08:OOh. Frequency difference. Subjects performed a forced-choice auditory recognition task that consisted of 1,051 trials in each of two sessions. They were instructed to press the button in one hand when they recognized the low tone and the button in their other hand when they recognized the high tone. Equal numbers of high and low tones were presented in each stimulus condition, in random order. Sessions were counterbalanced with respect to stimulus loudness and frequency differences. Each target tone lasted 100 ms. Following the termination of the target signal, the subject had 2500 ms to respond before the next trial was presented. A button press was followed by a 500ms intertrial interval. Failure to respond was scored as an error. Response times were calculated from the onset of the stimulus to the button-press time. Total time at the task varied according to how quickly the subject responded. Every subject completed two sessions. In Session A, the high tone was 1500 Hz; the low tone, 800 Hz (frequency difference = 700 Hz). In Session B, the high tone was 1500 Hz, and the low

tone, 1300 Hz (frequency difference = 200 Hz). The order of these sessions was randomized and counterbalanced. Before the actual testing, each patient practiced listening to pairs of high and low tones with a 700-Hz frequency difference until he or she was able to distinguish the difference between single pairs of tones at >90% accuracy. Tone presentation levels (volume). The presentation level changed after runs of 75 trials. Tone sets consisted of the following approximate levels: 83, 76, 72, and 69 dB. These lOO-ms stimuli were presented against a constant broadband noise background of 64 dB A scale. Presentation levels were determined from the sound-pressure levels measured with a B and K SLM 2203 meter, equipped with a ~-CC coupler and a B and K microphone. Decibel levels represent the energy associated with a lOO-mstone pulse. Actual soundpressure levels were measured with tones lasting 5000 ms. High and low frequency tones were randomly mixed within each level. The tones program was delivered and the responses were collected on a personal computer modified with an acoustic processor. Tones were presented binaurally through headphones. Tone characteristics. Pure tones that exhibited a symmetric harmonic content were used. At the onset and conclusion of each tone, there was a spread spectrum peak consistent with a switched sine wave. The power amplitude of each tone was measured at levels of 63, 25, and 6 dB, with the

78

H. H. Holcomb et al. /Psychiatry Research 57 (1995) 75-82

noise set to 0. At all levels, the two tones compared exactly, with better than 1-dB resolution. At 6 dB, the low tone was closer to the surrounding noise floor than the high tone. The noise spectrum measured at a middle to high level (40 dB) exhibited a flat distribution (f 5 dB) up to > 10 kHz; roll-off was pronounced after 16 kHz. Performance measures. Outcome measures were extracted from blocks of 25 consecutive trials. These block measures were as follows: mean accuracy, median response time for accurate trials, and percent of the incorrect trials in the block that fell within the slowest 25% of all incorrect responses found in all presentation levels in both sessions (slow errors). The last measure was used to calculate slow error response time. The distribution of slow errors was determined to show how changes in tone level were associated with shifts in decision speed and choice accuracy. The median was used to represent the response time in order to minimize the effect of outliers. 3. Results Figs. 1 and 2 demonstrate that for each frequency difference and at all tone levels, schizophrenic patients performed less accurately (F = 4.83, P = 0.0414) and more slowly (F = 37.35, P <

0.0001) than normal volunteers. For accuracy, there were significant main effects for frequency difference (F = 7.14, P = 0.0155) and presentation level (F = 3.84, P = 0.0302). As expected, accuracy dropped with the smaller frequency difference, but the pattern for presentation level was more complicated. When polynomial contrasts were applied to presentation level, a significant cubic trend (P = 0.0054) emerged. The frequency difference and presentation level main effects were not significant for median accurate response time. For slow errors, the main effect for presentation level (F = 3.71, P = 0.0373) and the frequency difference x presentation level interaction were significant (F = 3.14, P = 0.0488). For all of the dependent variables, there were no significant interactions involving group membership. The variability of the responses of the normal volunteers and the schizophrenic patients in all conditions, however, differed substantially. Across presentation levels, schizophrenic patients exhibited high levels of accuracy variance at each frequency difference: 700 Hz, SD = 133.83; 200 Hz, SD = 146.10. Normal volunteers displayed low accuracy variance levels: 700 Hz, SD = 33.45; 200 Hz, SD = 36.0. Attempts to transform the data to modify these differences in variability were not successful. However, analysis of variance, par-

90

60 69

72

76

83

Fig. 1. Percent. accuracy vs. tone presentation level. Frequency difference = 700 in Session A (left), 200 in Session B (right). Accuracy is greater for Session A compared with Session B, and in normal volunteers (nv) compared with schizophrenic subjects (sz). Post hoc analysis for volume level indicates a significant cubic trend. At the lowest presentation level, in Session 3, subjects in both groups exhibit greater accuracy than at intermediate presentation levels.

H. H. Holcomb et al. /Psychiatry Research 57 (1995)

75-82

79

Fig. 2. Median accurate response time vs. tone presentation level. Frequency difference = 700 in Session A (left), 200 in Session B (right). The response time is slower in schizophrenic patients (sz) than in normal volunteers (nv). The predicted slowing impact of lower presentation levels, independent of frequency difference and diagnostic group, is not significant (o = 0.14). Post hoc analysis of volume reveals a weak linear trend (a = 0. IO). As shown in Session B, sz subjects tend to be slower in their accurate response times at lower presentation levels: level 83 = 558 ms (mean), 76 = 567 ms, 72 = 586 ms, and 69 = 595 ms. This trend was weak (o = 0.10).

titularly with equal sample sizes, is extremely robust with respect to violations of homogeneity of variance. According to Harris (1985), normal curve based F tests can be considered valid if the ratio between the smallest and largest sample variance is < 20:l. The data for the schizophrenic and normal control subjects were also analyzed separately to

clarify their relative contribution to the main effect findings. The effect of the change in frequency difference from 700 to 200 Hz was a significant drop in accuracy from 89% to 80% for schizophrenic subjects (F = 6.14, P = 0.0327), while the change in accuracy for normal control subjects from 96% to 94% was not significant (F = 2.64, P = 0.1428). The changes in reaction time and slow errors for

Fig. 3. Percent slow errors vs. tone presentation level. Frequency difference = 700 in Session A (left), 200 in Session B (right). The distribution of slow error responses is significantly predicted by the volume, irrespective of diagnostic group or frequency ditference. Post hoc analysis reveals a linear, group-specific (normal volunteers, nv), volume-level-dependent distribution of slow errors in Session B. Schizophrenic subjects (sz) did not show this session-specific trend.

80

H.H. Holcomb et al. /Psychiatry Research 57 (1995) 75-82

frequency difference 200 versus 700 were not significant for either group (Fig. 3). The only significant effect of presentation level was the percent of slow errors for normal volunteers (F = 5.6, P = 0.0355). The examination of the polynomial contrasts for this effect demonstrated a cubic trend (F = 6.92, P = 0.0302). The same effect for schizophrenic patients suggested a similar trend (F= 3.11, P = 0.1099). 4. Discussion A forced-choice auditory discrimination task was used to compare the responses of normal volunteers and schizophrenic subjects to auditory S:N degradation. Through the modification of target tone volume and the frequency difference between the two tones, identical stimulus changes were presented to the two diverse diagnostic groups. Reductions in volume and frequency difference are well-established signal degradation maneuvers. We used these to determine whether the two groups would respond similarly to S:N manipulations, in spite of their scalar differences in response time and discrimination accuracy. Consistent with studies by Cardozo (1974) and Tanner (19&l), reductions in the frequency difference between two tones and reductions in the presentation level significantly reduced tonediscrimination accuracy in the schizophrenic group. As anticipated, we found patients to be slower and less accurate than normal volunteers in both sessions and across all presentation levels. The large variance within the schizophrenic group subverted our efforts to confirm an accuracy x diagnosis interaction. Nonetheless, schizophrenic patients exhibited a marked decline in accuracy with degraded stimuli, whereas normal volunteers did not. Both groups exhibited an unexpected pattern in their accuracy levels when plotted against volume levels. Fig. 1 (right panel) revealed a curvilinear (cubic) trend in accuracy versus tone level. When the presentation level was lowest, the subjects in both groups reversed an apparent downward trend and improved their accuracy. The causes for this improvement are unclear. It is possible that subjects overcompensated for the reduced presenta-

tion level by being more attentive. Replication of this phenomenon in future studies will help us to determine its stability and validity. It will be particularly helpful to measure the change in response time associated with reduced accuracy in normal volunteers when they are challenged with discrimination tasks that use small frequency differences. The post hoc analysis of slow error distribution suggests that schizophrenic patients are more likely than normal subjects to become particularly slow when trying to recognize tones that differ by only 200 Hz. When they performed the more difficult recognition task at low presentation levels, their incorrect choices were associated with increasing numbers of slow response times. Normal subjects showed no tendency to slow their response times when they were making incorrect choices, in Session B, at low presentation levels. It is, however, important to note the substantial differences in the error rates of patients and normal volunteers. In Session A, the normal volunteers had an average accuracy of 96%; schizophrenic subjects had an accuracy of 89%. But under reduced S:N conditions, the schizophrenic group fell to an average accuracy of 80% and the normal volunteers maintained an average accuracy of 94%. The large between-group accuracy differences in the two stimulus conditions reduce the interpretability of slow error patterns in the two groups. We expect that normal subjects would show a slow error pattern similar to that seen in schizophrenic subjects if they were challenged with low S:N and unable to confidently recognize the frequency of a target tone. These findings confirm that accurate tone discrimination, in schizophrenic subjects, is sensitive to signal degradation through frequency difference reduction. They also point up the marked group differences in performance between schizophrenic and normal individuals as measured by accuracy and response time. The linear change in responses of schizophrenic subjects to reduced volume levels in the more difficult session, where frequency difference = 200, highlights this group’s reduced processing capacity in the face of signal degradation. This was also found by Nestor et al. (1991). Small increments in task difficulty, associ-

H.H. Holcomb et al. /Psychiatry Research 57 (1995) 75-82

ated with a reduced S:N ratio, apparently challenge the internal resources available to the schizophrenic population (Nuechterlein and Dawson, 1984); those same increments do not challenge the normal volunteer group. Although this study directly supports a mode1 of reduced resource availability in schizophrenia (Callaway and Naghdi, 1982; Nuechterlein and Dawson, 1984; Granholm et al., 1991), it does not help us to understand why schizophrenic subjects are less accurate, slower, and more sensitive to S:N degradation. When used in conjunction with dynamic brain imaging, this and other experimental designs that track performance changes against controlled S:N manipulations may be helpful. Decisions that require greater resources, as are demanded by subtle or ambiguous stimulus differences, substantially modify neural activity in specitic brain regions (Spitzer et al., 1988; Spitzer and Richmond, 1991). By plotting physiological and behavioral changes against one another, an investigator may learn to what extent a set of interacting regions are necessary or sufficient to support a shift in resource demand. As demand changes, the physiological activity pattern appropriate for those particular responses should also change. Planned, systematic changes in an auditory stimulus frequency can be used to modify targetdiscrimination difficulty. In this study, schizophrenic patients were less accurate in the discrimination of tones that differed by 200 Hz than in the discrimination of tones that differed by 700 Hz (89% vs. 80%). Previous work by Tanner (1964) and Cardozo (1974) suggests that trained normal volunteers will also find tones more difficult to recognize when their frequencies are made similar (5- 100 Hz between levels of 1500-2000 Hz frequency). Individual subjects and diagnostic groups will differ in their abilities to recognize tones that vary in frequency differences. By manipulating this experimental variable, the investigator should be able to provide diverse subjects with a task that is nearly equivalent with respect to stimulus relevance or resource demand. To achieve reduced accuracy levels of 80-90%, normal subjects may require frequency differences of only 5- 15 Hz (Tanner, 1964) when using a

81

reference tone of 1500 Hz. On the basis of data presented here, schizophrenic patients may not be able to consistently discriminate tones that differ by < 200 Hz in order to maintain accuracy levels of 80-90%, under similar conditions. To some extent, the difficulty or demands of a task determine a subject’s brain-activity patterns. It is not known to what extent task demand modifies activity patterns throughout the brain. Schizophrenic patients and normal volunteers tend to differ in their capacity to make repetitive, difficult decisions. For this reason, investigators who attempt to determine brain-activity differences between these groups will benefit by studying them in similar states of task difficulty. By measuring brain blood flow or electrical current conduction patterns in subjects performing at similar levels of task difficulty, the investigator better controls for individual variations in task-related behavior. To the extent that these variations contribute to physiological activity patterns, it is advantageous to control them. This will improve the likelihood of discovering group functional differences that are independent of task performance. Acknowledgments Appreciation is expressed to Pamela Caudill for extensive editorial assistance, Jeffrey Sieracki for computer programming, Craig Formby for calibrating auditory presentation levels, the nursing staff of the MPRC Inpatient Research Unit for their assistance in the conduct of our studies, and Susan Nusbaum for assistance in preparing the manuscript. This work was supported by grants from the National Institute of Mental Health (MH-4421 l-04, MH-40279-05, and MH-42234-04). References Callaway, E. and Naghdi, S. (1982) An information processing model for schizophrenia. Arch Gen Psychiatry 39, 339-347. Cardozo, B.L. (1974) Facts and models in hearing. In: Zwicker, E. and Terhardt, E. (Eds.), Proceedings ofthe Symposium on Psychophysical Models and Physiological Facts in Hearing. Springer-Verlag, New York, pp. 164-177. Cohen, R., Semple, W.E., Gross, M., Nordahl, T.E., DeLisi, L.E., Holcomb, H.H., King, A.C., Morihisa, J.M. and Pickar, D. (1987) Dysfunction in a prefrontal substrate of

82

H.H. Holcomb et al. /Psychiatry Research 57 (1995) 75-82

sustained attention in schizophrenia. Life Sci 40, 203 I-2039. Cohen, R.M., Semple, W.E., Gross, M., Nordahl, T.E., Holcomb, H.H., Dowling, MS. and Pickar, D. (1988) The effect of neuroleptics on dysfunction in a prefrontal substrate of sustained attention in schizophrenia. Life Sci 43, 1141-1150. Goldberg, T.E., Weinberger, D.R., Pliskin, N.H., Berman, K.F. and Podd, M.H. (1989) Recall memory deficit in schizophrenia, a possible manifestation of prefrontal dysfunction. Schizophr Res 2, 251-257. Granholm, E., Asamow, R.F. and Marder, S.R. (1991) Controlled information processing resources and the development of automatic detection responses in schizophrenia. J Abnorm Psycho1 100, 22-30. Harris, R.J. (1985) A Primer of Multivariate Statistics. Academic Press, Inc., New York, pp. 331-334. Hyvarinen, J., Poranen, A. and Jokinen, Y. (1980) Influence of attentive behavior on neuronal responses to vibration in primary somatosensory cortex of the monkey. J Neurophysiol43,

870-882.

Kugler, B.T. and Caudrey, D.J. (1983) Phoneme discrimination in schizophrenia. Br J Psychiatry 142, 53-59. Meyer, E., Ferguson, S.S.G., Zatorre, R.J., Alivisatos, B., Marrett, S., Evans, A.C. and Hakim, A.M. (1991) Attention modulates somatosensory cerebral blood flow response to vibrotactile stimulation as measured by positron emission tomography. Ann Neural 29, 440-443. Mirsky, A.F. and Kometsky, C. (1968) The effect of centrally acting drugs on attention. In: Efron, D.H. (Ed.), Psychopharmacology A Review of Progress. (Public Health Service Publication 1836) Superintendent of Documents, U.S. Government Printing Office, Washington, DC. Moran, J. and Desimone, R. (1985) Selective attention gates visual processing in the extrastriate cortex. Science 229, 182-784.

Nestor, P.G., Faux, S.F., McCarley, R.W., Sands, SF., Horvath, T.B. and Peterson, A. (1991) Neuroleptics improve sustained attention in schizophrenia, a study using signal detection theory. Neuropsychopharmacology 4, 145-149. Nuechterlein, K.H. and Dawson, M.E. (1984) Information processing and attentional functioning in the course of schizophrenic disorder. Schizophr Bull IO, 160-203. Nuechterlein, K.H., Parasuraman, R. and Jiang, Q. (1983) Visual sustained attention, image degradation produces rapid sensitivity decrement over time. Science 220, 327-329. Overall, J.E. and Gorham, D.R. (1962) The Brief Psychiatric Rating Scale. Psycho/ Rep IO, 799-812. Richmond, B.J. and Sato, T. (1987) Enhancement of inferior temporal neurons during visual discrimination. J Neurophysiol 58, l292- 1306. Spitzer, H., Desimone, R. and Moran, J. (1988) Increased attention enhances both behavioral and neuronal performance. Science 240, 338-340, Spitzer, H. and Richmond, B.J. (1991) Task difficulty, ignoring, attending to, and discriminating a visual stimulus yield progressively more activity in inferior temporal neurons. Exp Brain Res 83(Z), 340-348.

Spohn, H.E., Lacoursiere, R.B., Thompson, K. and Coyne, L. (1977) Phenothiazine effects on psychological and psychophysiological dysfunction in chronic schizophrenics. Arch Gen Psychiatry 34, 633-644,

Tanner, W.P. (1964) Theory of recognition. In: Swets, J.A. (Ed.), Signal Detection and Recognition by Human Observers. John Wiley and Sons, New York, pp. 413-430. Weinberger, D.R., Berman, K.F. and Zec, R.F. (1986) Physiologic dysfunction of dorsolateral prefrontal cortex in schizophrenia: I. regional cerebral blood flow evidence. Arch Gen Psychiatry 43, I l4- 124.