Psychology

Vocal Expression in Schizophrenia: Less Than Meets the Ear
Alex S. Cohen and Kyle R. Mitchell Louisiana State University
Nancy M. Docherty Kent State University
William P. Horan VA Greater Los Angeles Healthcare System, University of California, Los Angeles
Abnormalities in nonverbal communication are a hallmark of schizophrenia. Results from studies using symptom rating scales suggest that these abnormalities are profound (i.e., 3–5 SDs) and occur across virtually every channel of vocal expression. Computerized acoustic analytic technologies, used to overcome practical and psychometric limitations with symptom rating scales, have found much more benign and isolated abnormalities. To better understand vocal deficits in schizophrenia and to advance acoustic analytic technologies for clinical and research applications, we examined archived speech samples from 5 separate studies, each using different speaking tasks (patient N 309; control N 117). We sought to: (a) use Principal Component Analysis (PCA) to identify independent vocal expression measures from a large set of variables, (b) quantify how patients with schizophrenia are abnormal with respect to these variables, (c) evaluate the impactofdemographicandcontextualfactors(e.g.,studysite,speakingtask),and(d)examinetherelationship between clinically—rated psychiatric symptoms and vocal variables. PCA identified 7 independent markers ofvocalexpression.Mostofthesevocalvariablesvariedconsiderablyasafunctionofcontextandmanywere associated with demographic factors. After controlling for context and demographics, there were no meaningful differences in vocal expression between patients and controls. Within patients, vocal variables were associatedwitharangeofpsychiatricsymptoms—thoughonlypauselengthwassignificantlyassociatedwith clinically rated negative symptoms. The discussion centers on explaining the apparent discordance between clinical and computerized speech measures.
General Scientific Summary Patients with schizophrenia are rated by clinicians as having profound and chronic vocal expressive deficits. Data from this study, using computerized vocal analysis of patients, suggests that vocal expression in patients is similar to nonpatients. Vocal deficits, if they occur in schizophrenia, are much more subtle than clinical ratings suggest, and may be contextually dependent.
Keywords: schizophrenia, negative, alogia, blunted affect, computer
Abnormalities in vocal expression, defined in terms of decreased/increased vocal production (i.e., alogia, pressured speech) and intonation/emphasis (i.e., blunted affect, affective lability) are a staple of schizophrenia and of serious mental illness more
generally (Cohen, Najolia, Kim, & Dinzeo, 2012; Galynker, Cohen, & Cai, 2000; Kulhara & Chadda, 1987; Mueser et al., 2010; Tremeau et al., 2005). Despite vocal expression reflecting an important Research Domain Criteria (RDoC; “Production of NonFacial Communication”) our understanding of its nature is poor. An obstacle in understanding and measuring vocal abnormalities is a reliance on interviewer-based rating scales (e.g., Andreasen, 1989; Kirkpatrick et al., 2011; Kring, Gur, Blanchard, Horan, & Reise, 2013; Lukoff, Nuechterlein, & Ventura, 1986). These scales are: relatively insensitive to change in a treatment context given their limited range of response options and ambiguous operational definitions, produce ordinal data that often cover wide temporal swaths, and are imprecise for isolating specific behaviors (Alpert, Shaw, Pouget, & Lim, 2002; Cohen, Alpert, Nienow, Dinzeo, & Docherty, 2008; Cohen & Elvevåg, 2014; Cohen, Mitchell, & Elvevåg, 2014; Mueser, Sayers, Schooler, Mance, & Haas, 1994; Stahl & Buckley, 2007). Emerging computerized technologies allow for assessment of vocal symptoms with near perfect “interrater” reliability and greater sensitivity and specificity for isolating specific behaviors than clinical rating scales, and thus, can provide
Editor’s Note. David B. Watson served as the Guest Editor for this article.—SHG
Alex S. Cohen and Kyle R. Mitchell, Department of Psychology, Louisiana State University; Nancy M. Docherty, Department of Psychology, Kent State University; William P. Horan, VA Greater Los Angeles Healthcare System, University of California, Los Angeles. This project was funded by Grant R03MH092622 from the National Institute of Mental Health to AC. We thank the patients and subjects who volunteered their time to participate in the study. We are also grateful to Matt Calamia for his statistical advice and consultation. Correspondence concerning this article should be addressed to Alex S. Cohen, Department of Psychology, Louisiana State University, 236 Audubon Hall, Baton Rouge, LA 70803. E-mail: acohen@lsu.edu
ThisdocumentiscopyrightedbytheAmericanPsychologicalAssociationoroneofitsalliedpublishers. Thisarticleisintendedsolelyforthepersonaluseoftheindividualuserandisnottobedisseminatedbroadly.
Journal of Abnormal Psychology © 2016 American Psychological Association 2016, Vol. 125, No. 2, 299–309 0021-843X/16/$12.00 http://dx.doi.org/10.1037/abn0000136
299
a more fine grained, and objective analysis compared with clinical rating scales assessing vocal psychopathology (Cohen & Elvevåg, 2014). These technologies are particularly attractive in that they can be used in mobile assessment and telemedicine applications, and can be automated such that they are efficient and inexpensive for clinical use (e.g., Satt, Hoory, König, Aalten, & Robert, 2014). Moreover, vocal variables have been associated with clinical measures of functioning in at least some studies (Cohen, Najolia, Kim, & Dinzeo, 2012), and the behaviors they tap are part of the operational definition of social competence (e.g., Bellack, Morrison, Mueser, Wade, & Sayers, 1990). The present study applied acoustic-analytic technologies to understand vocal deficits in schizophrenia using archived natural speech samples from five independent studies. A recent meta-analysis (Cohen, Mitchell, et al., 2014) highlights potential limitations with clinical ratings scales and our poor understanding of vocal abnormalities in schizophrenia more generally. With respect to vocal expression ratings from the Scale for the Assessment of Negative Symptoms (e.g., blunted vocal affect; Andreasen, 1989), patients with schizophrenia versus nonpatient controls showed profound deficits on the order of 3 to 5 SDs (i.e., Cohen’s d 3.42–4.39). In contrast, computer-based measures of vocal expression were associated with more modest, and variable, abnormalities in patients. For example, patients showed normal vocal patterns in some variables (e.g., latency of response; d .21, k 2) and medium to large effects in variables related to vocal production (e.g., number of words expressed; d .60, k 6; average pause length, d 1.10; k 4). These findings suggest that, in contrast to results from symptom-rating based studies, only isolated aspects of vocal expression are abnormal in schizophrenia. Consider further a recent meta-analysis of six studies using behavioral coding of patients while engaged in structured laboratory tasks. The range of effect sizes was fairly broad (range of d=s .61 to 1.95; dmean 1.10; Hoekert, Kahn, Pijnenborg, & Aleman, 2007), perhaps reflecting differences in coding systems and tasks across studies. Moreover, ratings were fairly holistic in their appraisal of vocal expression; a potential issue in that vocal expression is multifaceted. Regardless, even behavioral ratings fail to approximate the magnitude of deficits associated with symptom rating scales. Localizing and clarifying the vocal abnormalities associated with schizophrenia is an important step toward understanding the nature and pathophysiology of negative symptoms, and for identifying biomarkers for their objective assessment. This was the first aim of this study. When evaluating vocal deficits in patients, it is important to consider that vocal expression is not static. In the general population, vocal expression is highly dynamic across contexts and is modulated by a number of factors, such as presence/absence of social stimuli (Wells & MacFarlane, 1998), cognitive resources (Cohen, Dinzeo, Donovan, Brown, & Morrison, 2015), and affective/arousal state (e.g., angry, happy; Cohen, Minor, Najolia, & Hong, 2009; Johnstone et al., 2007; Sobin & Alpert, 1999) to name a few. In schizophrenia, vocal deficits are conceptualized as being temporally stable and occurring across contexts (e.g., Kirkpatrick & Galderisi, 2008)—a phenomenon that has been referred to as “Affective Nonresponsivity” (Andreasen, 1989). Indeed, symptom rating scales tapping vocal deficits show modest to high levels of stability over extended periods of time (Long & Brekke, 1999; Strauss, Harrow, Grossman, & Rosen, 2010). However, several
studies using vocal/speech analysis while manipulating speaking context—by increasing the cognitive demands of the speaking task—found that aspects of vocal expression worsen as a function of context (i.e., word count in Barch & Berenbaum, 1994; pause length in Cohen, McGovern, Dinzeo, & Covington, 2014). This suggests that vocal deficits are exacerbated by certain contextual demands,andarenotasstaticasindicatedbysymptomratingscale data. Insofar as cognitive demands play a role, a wide range of contextual variables (presumably related to cognitive resources when speaking), such as complexity of speech topic, competing tasks (e.g., driving, walking) and familiarity with speaker may be important to understanding vocal deficits. This topic has received very little attention to date. Answering the question “Does vocal expression in schizophrenia vary as a function of contextual variables” was our second aim. Additionally, symptom heterogeneity and comorbidity in schizophrenia may be relevant. Vocal abnormalities in schizophrenia have primarily been examined in the context of negative symptoms, for which increased pauses, decreased signal variability (e.g., change in frequency, intensity), and tonal abnormalities have been associated with negative symptom severity in at least some studies (Alpert et al., 2002; Cohen et al., 2008; Cohen, Mitchell, et al., 2014). However, these same vocal anomalies have been associated with other clinical disorders and symptoms, such as those involving depression and anxiety (Cannizzaro, Harel, Reilly, Chappell, & Snyder, 2004; Cohen, Kim, & Najolia, 2013; Mundt, Vogel, Feltner, & Lenderking, 2012); symptoms that are common in schizophrenia and conceptually related to negative symptoms. Moreover, vocal exaggerations, as opposed to deficits, may be observed in some patients with schizophrenia in that hostility, mania, and even thought disorder are conceptually related to increases in vocal production (e.g., pressured speech), intonation, and emphasis (Sobin & Alpert, 1999). The third aim of this study was to evaluate relationships between vocal expression and a broad range of psychiatric symptoms. The present study leveraged acoustic analytic technologies to understand vocal expression in schizophrenia and nonpsychiatric controls from a large database of speech samples, collected from five different studies—each using different speaking tasks and procedures. Given the relative novelty of this endeavor, and the considerable inconsistency in vocal features used in prior studies (Cohen, Mitchell, et al., 2014; Cohen, Renshaw, Mitchell, & Kim, 2015), our first step was to subject a relatively large number of commonly used vocal variables to data reduction procedures (e.g., Principal Component Analysis [PCA]). Following this, we evaluated the degree to which vocal expression varied across contextual factors (represented in the five separate speaking tasks/study sites) and whether, after controlling for these contextual factors, patients differed from controls. Finally, we evaluated the links between vocal expression and various psychiatric symptoms in the patient group to explore potential heterogeneity across patients.
Method
Participants
Data were aggregated from five separate archived studies (Cohen et al., 2008; Cohen et al., 2013; Cohen, McGovern et al., 2014; Docherty, 2012; Horan et al., 2009) conducted with outpatients
ThisdocumentiscopyrightedbytheAmericanPsychologicalAssociationoroneofitsalliedpublishers. Thisarticleisintendedsolelyforthepersonaluseoftheindividualuserandisnottobedisseminatedbroadly.
300 COHEN, MITCHELL, DOCHERTY, AND HORAN
and long-term forensic inpatients meeting criteria for Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition (DSM–IV; American Psychiatric Association, 2013) schizophrenia or schizoaffective disorders. Diagnoses were established using structured clinical interviews (First, Gibbon, Spitzer, & Williams, 1996). All patients were deemed clinically stable at the time of testing and were receiving treatment from multidisciplinary teams. Nonpatient controls from these studies were also examined, for whom lack of psychosis and mood-spectrum diagnoses was established by structured clinical interviews (First et al., 1996). Data for 309 patients and 117 nonpatient controls were available. Table 1 contains demographic and clinical information. Each of the studies was approved by the appropriate Institutional Review Boards and all subjects provided written informed consent before beginning the study.
Diagnostic and Symptom Ratings Psychiatric symptoms were measured using either the 18-item (Overall & Gorham, 1962) or the Expanded Brief Psychiatric Rating Scale (BPRS; Lukoff et al., 1986). Unfortunately, only factor scores (per Guy, 1976) were available for all five studies. Cronbach values were generally adequate (i.e., .60), though two scales (i.e., Activation and Hostility) were combined to achieve this. Factors included: Thought Disturbance (.72), Anergia(.73),Anxiety-Depression(.74.),andActivation/ Hostility (.62).
Speaking Tasks (i.e., Context) Across studies, participants were asked to produce speech as part of one of five different tasks that varied in topic, cognitive demands, interviewer participation and length (see Table 1). For Studies 1 (Cohen et al., 2008) and 5 (Docherty, 2012), research assistants engaged the subject in a semistructured conversation about daily activities, hobbies, and interests. The speech sample for Study 1 was 300 s and 600 s for Study 5; though because of computing demands of processing extended speech, only the first 300 s of the latter samples were analyzed. Studies 2 (Cohen et al., 2013) and 3 (Cohen, McGovern et al., 2014) involved having
subjects provide monologues on daily activities/hobbies (180 s) and neutrally valenced autobiographical memories (90 s) respectively, for which the interviewer was absent the subjects’ field of vision and was silent for the task. Although Studies 2 and 3 were conducted in similar geographic regions and similar treatment settings and with similar research assistants, we opted to consider them separately given differences in speaking time and content, which in a prior study of these tasks in nonpatients was associated with significant differences in vocal variables (Cohen, Renshaw, Mitchell, & Kim, 2015). Study 4 (Horan et al., 2009) involved verbal responses from a social cognition task, involving responses to 12 different videos containing animated shapes (range of responses 42–75 s; Castelli, Frith, Happe, & Frith, 2002). All interviewer speech was digitally removed from the speech samples before analysis. Age did not significantly vary across the five studies (ps .68). Study 5 had a significantly higher percentage of women than the other studies (2[4] 13.38, p .01). With respect to ethnicity, Study 3 had significantly more White than the other studies (2[4] 12.64, p .01), and Studies 3 and 4 had significantly fewer Blacks than the other studies (2[4] 37.97, p .001). Some significant differences in Thought Disturbance (Study 5 all other studies), Anxiety Depression (Study 1 all other studies), and Activation-Hostility (Study 5 than Studies 1, 3, and 4) were observed as well (Scheffé test ps .01).
Computerized Acoustic Analysis of Natural Speech The Computerized assessment of Affect from Natural Speech (CANS; Cohen, Hong, & Guevara, 2010; Cohen et al., 2015), developed by our lab to assess vocal expression from natural speech, was used here. The CANS is based, in part, on Praat (Boersma & Weenink, 2013), a program that has been used extensively in speech pathology and linguistic studies. Digital audio files were organized into “frames” for analysis (i.e., 100 per second). During each frame, four basic speech properties were quantified: Fundamental Frequency (F0), 1st Formant (F1), 2nd Formant (F2), and Intensity (see Figure 1). To limit potential background noise, we used optimization filters for measuring F0 (Vogel, Maruff, Snyder, & Mundt, 2009): Low: 75 Hz; High: 300 Hz. Because of the nonlinear nature of the hertz frequency scale,
Table 1 Descriptive and Study Information
Study1a Study 2b Study 3c Study 4d Study 5e
N controls/patients 21–65 0–31 33–42 40–53 23–118 Patient setting Forensic–inpatient Outpatient Outpatient Outpatient Outpatient % Female (Con–Pts) 33–28% 0–36% 52–31% 25–25% 48–44% Age controls 38.33 10.75 n/a 40.88 12.24 39.73 9.23 38.26 9.17 Patients 42.06 8.44 39.68 10.76 42.33 12.02 40.30 10.71 40.13 7.36 % White (Con–Pts) 57–35% 0–32% 49–52% 40–32% 39–33% % Black (Con–Pts) 43–65% 0–68% 42–43% 30–32% 61–64% BPRS factor scores Thought disturbance 2.80 1.29 2.60 1.15 2.79 1.37 2.54 1.19 1.79 .74 Anxiety/depression 1.50 .46 2.67 1.16 2.25 1.16 2.38 .82 2.62 1.14 Anergia 1.63 .70 1.98 .98 1.63 .84 1.67 .65 1.56 .58 Activation-hostility 2.80 1.29 2.60 1.15 2.79 1.37 2.54 1.19 1.79 .74 Speaking task Conversation Monologue Monologue Cognitive task Conversation Speaking task length 300 s 180 s 90 s 42–75 s 300 s a Cohen et al., 2008. b Cohen et al., 2013. c Cohen, McGovern, et al., 2014. d Horan et al., 2009. e Docherty, 2012.
ThisdocumentiscopyrightedbytheAmericanPsychologicalAssociationoroneofitsalliedpublishers. Thisarticleisintendedsolelyforthepersonaluseoftheindividualuserandisnottobedisseminatedbroadly.
301VOCAL ANALYSIS AND SCHIZOPHRENIA
F0, F1, and F2 values were converted to semitones—a linear scale used for parametric statistics of hertz-scale data. There is no consensus on how to quantify vocal signals, and there is considerable variability in summary statistics used (e.g., means, SDs) and across variable time scales (e.g., computed across the whole vs. portions of the speech sample). To address this issue, we identified 21 systematically defined and commonly reported variables, based in part on a recent psychometric evaluation of 1,500 young adults (Cohen et al., in press) and a broader literature review(Huttunen,Keranen,Vayrynen,Paakkonen,&Leino,2011; Scherer, 2003; Sobin & Alpert, 1999; Table 2). Three of these variables involved “Vocal Production,” defined in terms of pauses (i.e., 50 ms or more of signal absence) bounded by an epoch of F0 signal greater than 150 ms in length with no contiguous pause greater than 50 ms. The remaining measures were based on variability in F0, F1, F2 and intensity signals in terms of three different epochs: “local,” “global,” and “instability.” Local refers to variability within utterances, whereas global refers to variability across utterances (i.e., across the speech sample). We computed local and global measures for both “Range” (i.e., the difference between the highest and lowest values within utterances) and for SD scores. For
F0andintensitysignals,wealsocomputedinstability,whichrefers tosubtlesignalfluctuations(i.e.,changeinconsecutiveframes;see Table 2).
Analyses The analyses were conducted in four steps. Our first step focused on data reduction of the 21 vocal variables using PCA with Oblimin rotation and Kaiser normalization. A fixed factor number was determined using parallel analysis (O’Connor, 2000). Our primary goal was to find a common factor solution for patients and controls to facilitate comparison between them; so PCA were first conducted on all participants simultaneously and then verified in patients and controls separately. Component factor scores were computed and used in subsequent analyses. Second, hierarchical linear regressions were used to evaluate the relative relationships between the seven vocal variables (dependent variables) and diagnostic group (Step 3) above and beyond those associated with demographic variables (age, sex, and ethnicity; Step 1) and study (Step 2). The ethnicity (2 separate variables) and study (5 separate variables) were “dummy-coded.” To complement these analyses,
Figure 1. Vocal properties examined in this study.
ThisdocumentiscopyrightedbytheAmericanPsychologicalAssociationoroneofitsalliedpublishers. Thisarticleisintendedsolelyforthepersonaluseoftheindividualuserandisnottobedisseminatedbroadly.
302 COHEN, MITCHELL, DOCHERTY, AND HORAN
bivariate correlations between all predictor and independent variables were computed. Third, we evaluated the relationships between psychiatric symptoms and vocal variables using both Pearson and partial (i.e., controlling for age, sex, and study) correlations. Finally, we complemented the analyses in Step 3 by comparing vocal variables between patients with varying levels of clinically rated negative symptoms, including: No (i.e., all Anergia symptoms rated as “absent”; 90), Mild (i.e., average of Anergia ratings in the “mild” range [1 to 3]; 179), and High (i.e., average of Anergia ratings greater than “mild” [greater than 3]; 23). For normalization purposes, “extreme” vocal scores (i.e., 3.5 SDs) were “winsorized” to 3.5.
Results
Data Reduction of Vocal Variables
Parallel analysis of the 21 vocal variables in the entire sample suggested a five-factor solution. The five-factor PCA solution of these variables explained 79.15% of the variance, though the factor structure did not replicate separately in the patient (80.73% of the variance) and control groups (79.41% of the variance). In the patient sample, the instability items cohered heavily together (i.e., component scores .60) whereas in the control sample they did not; instead loading heavily with three of the four Vertical Tongue Movement items (i.e., component scores .51). In the patient sample, Pause N did not load heavily with any of the factors (i.e., component scores .32) whereas in controls, Pause N heavily loaded with other Pause items and a single F1 item (i.e., component scores .60). Excluding Pause N, Pitch Instability, and Volume Instability from the PCA resulted in a coherent four-factor solution (as indicated by Parallel analysis) that was consistent across both patient and control samples. The PCA solution explained 85.16% of the variance. The KMO (i.e., .61) and Bartlett’s
(p .0001) test statistics were both adequate for this analysis. The pattern matrix for patients and controls are presented in Table 3.A component correlation matrix of the four factors suggested they were not redundant with each other (e.g., mean r values were .13 for controls and .11 for patients; rs .36; Table 4), nor were they redundant with the three variables excluded from the PCA. The component scores for each factor for the patient and control groups were highly correlated with each other (rs .92). Given the potential independence of the Instability Scores in the control groups, their relatively low correlation with each other (e.g., r .27 for patients), and the fact that a prior study of these variables revealed they are independent with respect to some psychopathological processes (Cohen et al., in press), we decided to examine these variables separately. Additionally, we examined Pause N separately. Consequent analyses focused on seven variables.
Vocal Variables: Relative Contributions of Demographic, Context, and Diagnostic Group Variables Bivariate correlations (see Table 5) and hierarchical linear regressions (see Table 6) between predictor and independent variables were computed. Regressions indicated that significant contributions were made by demographic (i.e., three of seven regressions) and contextual (seven of seven regressions) variables. Diagnostic group was negligible for all seven analyses and in all but one analysis, were not statistically significant. The contribution of context was quite pronounced, accounting for over half of the variance in some variables. Inspection of the regression coefficients suggested that women had shorter Pauses ( .10, t 2.20, p .03) and greater Intonation ( .24, t 5.10, p .001) and Tongue Movement ( .11, t 2.25, p .03). Age was associated with shorter Pauses ( .11, t 2.39, p .02) and less Pitch Instability ( .25, t 5.33, p .001). Whites
Table 2 Summary of Vocal Variables Included in This Study
Pauses
Pauses M – Mean of pause lengths (ms). Pause SD – SDof pause lengths. Pause N – Total N of pauses.
Intonation Vertical tongue movement F0 SD global – SD of all SDs computed within utterances. F1 SD global – SD of all SDs computed within utterances. F0 SD local – M of all SDs computed within utterances. F1 SD local – M of all SDs computed within utterances. F0 range global – SD range computed within utterances. F1 range global – SD range computed within utterances. F0 range local – M range computed within utterances. F1 range local – M range computed within utterances.
Sagittal tongue movement Emphasis F2 SD global – SD of all SDs computed within utterances. INT SD global – SD of all SDs computed within utterances F2 SD local – M of all SDs computed within utterances. INT range global – SD range computed within utterances. F2 range global – SD range computed within utterances. INT range local – M range computed within utterances. F2 range local – M range computed within utterances. INT SD local – M of all SDs computed within utterances.
Pitch instability Volume instability Pitch instability – Absolute value of F0 change in consecutively voiced frames within utterance (M across utterances). Volume instability – Absolute value of INT change in consecutively voiced frames within utterance (M across utterances). Note. INT Intensity.
ThisdocumentiscopyrightedbytheAmericanPsychologicalAssociationoroneofitsalliedpublishers. Thisarticleisintendedsolelyforthepersonaluseoftheindividualuserandisnottobedisseminatedbroadly.
303VOCAL ANALYSIS AND SCHIZOPHRENIA
versus non-Whites had shorter Pauses ( .28, t 3.69, p .001) and greater Tongue Movement ( .26, t 3.18, p .002). To illustrate the relationship between vocal variables and context, standardized means (both patients and controls; controlling for demographic variables) for each vocal variable were computed and plotted in Figure 2. There was no obvious pattern of differences across studies, though study five was relatively low in Tongue Movement and Emphasis whereas Study 1 was notably high in Intonation and Emphasis. Patients and controls were then compared in vocal variables for each study independently. Of the potential 28 analyses (seven variablesforfourstudies;Study2containednocontrols),onlyfour significant speech deficits were observed. In Study 1, patients versus controls showed significantly fewer Pauses (d .50, 95% confidence interval [CI; .72, .29]) and less Volume Instability (d .52, 95% CI [.73, .30]). In Study 3, patients versus controls showed less Intonation (d .59, 95% CI [.81, .37]), and in Study 4, patients showed longer Pauses (d .47, 95% CI [.68, .25]; all ps .05).
Vocal Variables and Psychiatric Symptoms Table 7 contains the correlations between symptom ratings and vocal variables. Despite some significant correlations, the magnitudes were all, with one exception, in the small to negligible range (i.e., rs .30). That being said, there were some statistically significant findings even after controlling for demographic and contextual variables. Anergia, but not any other variables, was significantly associated with increased Pause lengths. Depression/ anxiety symptoms were associated with fewer Pauses, and less Pitch Instability. Thought disorder was also associated with fewer Pauses, and also greater Intonation and Pitch Instability. Activation/Hostility was associated with greater Tongue Movement.
Clinically Defined Negative Symptoms The groups differed significantly only in Pause Length, F(3, 405) 3.30, p .02, which was driven by the Non-Negative group showing shorter pauses than the other groups (ps .03). The High-Negative group significantly differed from the NonNegative but not the Low-Negative group. Trend-level group differences were observed with Tongue Movement, F(3, 405) 2.31, p .08; which reflected the High-Negative group showing significantly more Tongue Movement than the Control and MildNegative symptom group (ps .03) and at a trend level compared with the No-Negative symptom group (p .08). Standardized scores are reported in Figure 3.
Discussion This study is the largest investigation of vocal abnormalities in schizophrenia using computerized analytic technologies to date. There were four main findings. First, our data reduction analysis identified seven independent vocal factors/variables. Given the overwhelming number of vocal variables available for analysis and
Table 3 Principal Components Analysis Results; Pattern Matrix for Patients and Controls (in Parentheses)

Tongue Emphasis Intonation Pause M
F1 range global .95 (.81) .01 (.02) .01 (.21) .07 (.01) F1 range local .92 (.86) .01 (.00) .10 (.13) .07 (.07) F2 range local .89 (.86) .01 (.01) .06 (.01) .13 (.30) F1 SD global .89 (.79) .01 (.02) .13 (.19) .10 (.13) F2 SD local .89 (.85) .00 (.03) .03 (.09) .16 (.32) F2 range global .87 (.77) .02 (.06) .14 (.13) .09 (.10) F1 SD global .86 (.64) .02 (.01) .03 (.08) .12 (.36) F2 SD global .78 (.68) .00 (.01) .16 (.23) .12 (.08) INT SD local .01 (.00) 1.00 (1.00) .01 (.00) .00 (.01) INT SD global .02 (.00) 1.00 (1.00) .01 (.00) .00 (.00) INT range global .01 (.00) 1.00 (1.00) .00 (.00) .00 (.00) INT range local .03 (.02) 1.00 (1.00) .02 (.01) .00 (.01) F0 SD global .21 (.25) .02 (.02) .92 (.87) .01 (.04) F0 range global .02 (.01) .00 (.01) .92 (.87) .01 (.10) F0 SD local .10 (.10) .00 (.01) .91 (.88) .05 (.20) F0 range local .23 (.30) .02 (.01) .83 (.80) .06 (.17) Pause SD .01 (.02) .03 (.08) .03 (.08) .96 (.90) Pause M .03 (.01) .02 (.02) .04 (.06) .95 (.92) Note. Values .35 are bolded. a The instability and pause N measures are presented independently (see text for elaboration).
Table 4 Zero-Order Correlation Matrix of Factor Scores for Controls (Italicized) and Patients
1 2 3 4 5 6 7
1. Pause M .04 .03 .01 .08 .10 .15 2. Pause N .28 .12 .14 .23 .02 .14 3. Intonation .14 .04 .06 .20 .02 .36 4. Pitch instability .14 .05 .04 .18 .00 .27 5.
Tongue .07 .23 .15 .31 .04 .07 6. Emphasis .18 .05 .01 .07 .02 .02 7. Volume instability .25 .07 .26 .21 .09 .01 p .05.
ThisdocumentiscopyrightedbytheAmericanPsychologicalAssociationoroneofitsalliedpublishers. Thisarticleisintendedsolelyforthepersonaluseoftheindividualuserandisnottobedisseminatedbroadly.
304 COHEN, MITCHELL, DOCHERTY, AND HORAN
the considerable variability in scores reported across studies (Cohen, Mitchell, et al., 2014), these findings provide a psychometrically informed framework for future vocal analysis in clinical populations. Second, vocal expression varied considerably across the studies examined in both patients and controls, indicating that contextual variables (e.g., speaking task, setting) should be considered when evaluating vocal abnormalities in schizophrenia. Third, when controlling for demographic and contextual variables, there were no meaningful differences between controls and patients in any of the variables examined. This finding is in stark contrast to the dramatic differences in symptom rating scales reported between patients, as a group, and controls (Cohen, Mitchell, et al., 2014). Even in patients with high levels of clinically rated negative symptoms, most facets of vocal expression were not abnormal (with the possible exception of pause length; discussed below). Fourth, vocal variables were significantly correlated (albeit at a small effect size level) with psychiatric symptoms in the patient group. The most important finding from this study concerns the lack of patient vocal deficits from computerized measures; remarkable because schizophrenia is defined, in part, based on negative symp
toms (i.e., alogia, blunted affect) and is associated with profound symptom rating scale abnormalities (Cohen, Mitchell, et al., 2014). We see three possible explanations for the discrepancy between clinical and computer measures. First, it could be that patient’s vocal deficits are actually much more benign in magnitude and limited in breadth than clinician’s ratings and diagnoses would suggest. Supporting this notion, evidence from our recent metaanalysis (Cohen, Mitchell, et al., 2014) suggests that only variables related to speech production (e.g., Pauses) are consistently and notably (i.e., d .80) abnormal in schizophrenia. Evidence of clinicians imprecision in evaluating patient speech has been reported in a line of research conducted by Murray Alpert and colleagues. Of particular note, Alpert, Pouget, and Silva (1995) manipulated the pause times of prerecorded patient speech, but no other acoustic characteristics, and asked clinicians to evaluate the recordings using symptom rating scales. Ratings of blunted vocal affect were inappropriately inflated, raising the question of whether measures of vocal expression (and perhaps negative symptoms more generally) are unduly anchored in patient’s pause times. Studies by Alpert and colleagues and our lab have found that clinical ratings of flat affect are significantly correlated with acoustically derived pause measures (but not measures of vocal affect) as well as global negative symptom scores, neuropsychological functioning and extrapyramidal side effects (Allan, Sison, Alpert, Connolly, & Crichton, 1998; Alpert, Kotsaftis, & Pouget, 1997; Alpert, Pouget, Welkowitz, & Cohen, 1993; Alpert, Rosen, Welkowitz, Sobin, & Borod, 1989; Alpert et al., 2002; Cohen et al., 2008, 2013). Accordingly, Alpert and colleagues have proposed that clinical ratings of vocal deficits are conflated by perceptions of global impressions of patient behavior, based on inferences from highly salient and potentially comorbid or confounding influences—a sort of “halo” effect. That being said, pauses were not abnormal in the present patient group. This leads us to a second potential explanation for the discordance between computerized and symptom-rating measures—namely that vocal deficits are dynamic and emerge only in certain contexts. Several lines of evidence support this claim. First, in the present study, nearly all vocal variables differed as a function of speaking task, setting or other contextual factors.
Table 5 Bivariate Correlations Between Vocal Variables and Demographic, Contextual, and Group Predictor Variables
Pause M Pause N Intonation
Pitch instability
Tongue Emphasis
Volume instability
Demographic variables Age .12 .05 .04 .24 .08 .07 .02 Sex .11 .05 .24 .11 .11 .04 .01 White ethnicitya .00 .10 .02 .06 .02 .06 .00 Blacka .13 .10 .01 .02 .01 .09 .04 Contextual variables Study 1a .02 .03 .37 .36 .52 .05 .46 Study 2a .07 .05 .13 .17 .20 .04 .39 Study 3a .23 .38 .08 .42 .28 .07 .48 Study 4a .55 .10 .00 .14 .01 .27 .13 Study 5a .28 .40 .18 .25 .76 .11 .09 Group variables Groupa .02 .15 .10 .07 .04 .12 .01 a Variable is dummy-coded. p .05.
Table 6 Relative Contributions of Demographic (Step 1), Contextual (Step 2), and Diagnostic (Control, Patient) Variables to Vocal Variables
Dependent variables
Demographica Contextualb Diagnosis
R2
F
R2
F
R2
F
Pauses M .07 7.79 .28 44.74 .01 5.59 Pause N .01 1.57 .23 32.63 .00 1.62 Intonation .06 6.73 .18 25.48 .00 1.94 Pitch instability .09 10.53 .33 58.11 .00 .17
Tongue .02 2.16 .72 282.24 .00 1.14 Emphasis .01 1.56 .07 7.97 .00 2.19 Volume instability .01 .61 .52 112.36 .00 .00 a Ethnicity, sex, and age. b Variables were dummy-coded and entered in a single step. p .05.
ThisdocumentiscopyrightedbytheAmericanPsychologicalAssociationoroneofitsalliedpublishers. Thisarticleisintendedsolelyforthepersonaluseoftheindividualuserandisnottobedisseminatedbroadly.
305VOCAL ANALYSIS AND SCHIZOPHRENIA
Second, patients showed abnormal pauses in several studies examined here suggesting that, in some but not all conditions, pauses may become abnormally exaggerated or increased in number. Third, there is evidence that cognition is a “contextual” variable that may underlie vocal deficits. Vocal characteristics are determined, in part, by the involvement and integration of a range of basic cognitive resources. Even in nonpatient populations, restricting these resources leads to more sparse speech and less variability (Barch & Berenbaum, 1994; Cohen, Dinzeo et al., 2015). Given that schizophrenia is associated with a broad range of cognitive deficits, (Cohen et al., 2007; Heinrichs & Zakzanis, 1998) it stands to reason that even minor strains on cognitive resources may have dramatic effects on vocal communication. Note that one of the speaking tasks where patients were abnormal in pause time involved a social cognition task that was at least somewhat demanding in cognitive resources. Finally, Alpert et al. (1997) found that certain pauses, namely those occurring between clauses and when “switching turns” during dyadic exchanges, were particularly pathognomonic in schizophrenia. In summary, the present findings are not inconsistent with the notion that abnormal pause times (a.k.a. alogia) may be limited to contexts for which cognitive resources are limited. Beyond cognitive variables, it is also possible that acoustic differences would have been more pronounced if the speech content was emotionally charged in some manner. With
respect to symptom rating scales, it may be that symptom raters may be implicitly weighting (or overweighting) these incidences when making their ratings. A third possibility is that symptom measures accurately capture vocal deficits that are somehow missed using the vocal measures used in the present study. In defense of the present measures, they are conceptually linked to clinical definitions of vocal deficits in schizophrenia and are commonly used in studies of clinical populations and in communication sciences more generally (Cohen, Dinzeo et al., 2015; Cohen et al., 2009; Johnstone et al., 2007; Sobin & Alpert, 1999; Sobin & Alpert, 1999; Wells & MacFarlane, 1998). Thus, they have shown convergent validity in understanding a broad range of important communicative functions. From a vocal perspective, it is hard to believe, for example, that “blunted vocal affect” would not involve less variability of F0 and Intensity signals within utterances. That being said, there are literally thousands of ways to process, organize, and quantify the human voice, and it is possible that an as-yet unidentified vocal characteristic will meaningfully differentiate schizophrenia from nonpatients. It may also be the case that vocal deficits can only be understood within the confines of dyadic exchanges. This is a particularly salient point in light of a recent study that patients with schizophrenia fail to match their speaking partners in vocal “synergy”—definedintermsofvocalmodulationsonenormallymakes during dyadic interaction (Dombrowski, McCleery, Gregory, & Docherty, 2014). It may be the case that vocal abnormalities require grounding within the contextual demands of their conversational partner, and are not abnormal in-and-of themselves. The Pitch Instability measure warrants further consideration, as it was correlated (albeit at a small effect size level) with nonnegative symptom factors (e.g., thought disturbance, anxietydepression). In terms of psychiatric correlates, Pitch Instability (vis a vis “jitter”) has been associated with stressor-provoked anxiety (Mendoza&Carballo,1998),buthasreceivedlittleattentioninthe schizophrenia literature. Pitch Instability reflects “microinstability” in vocal cord vibrations, or random modulation of the source signal. The psychological underpinnings of it are not well known, but there is evidence that they are influenced by arousal levels (vis a vis cardiovascular functions). Indeed, a recent double-blind placebo controlled study found that propranolol increased both cardiovascular functions and vocal jitter levels in individuals undergoing a laboratory stressor task (Giddens, Barron, Clark, &
Table 7 Correlations Between Clinically Rated Symptoms and Vocal Variables, Computed Separately Without Control, and Controlling for Sex, Ethnicity, and Context (in Parentheses) as Partial Correlations (Patients Only)
Thought disorder
Anxietydepression Anergia
Activationhostility
Pauses M .11 (.05) .02 (.07) .13 (.14) .06 (.09) Pause N .04 (.13) .15 (.15) .05 (.02) .20 (.10) Intonation .17 (.15) .10 (.08) .08 (.07) .01 (.09) Pitch instability .24 (.20) .01 (.13) .05 (.02) .02 (.11)
Tongue .34 (.00) .28 (.02) .03 (.10) .17 (.23) Emphasis .06 (.09) .03 (.02) .08 (.08) .10 (.07) Volume instability .04 (.04) .19 (.07) .14 (.08) .02 (.00) p .05.
Figure 2. Standardized means of vocal variables plotted as a function of study (combined patients and controls).
ThisdocumentiscopyrightedbytheAmericanPsychologicalAssociationoroneofitsalliedpublishers. Thisarticleisintendedsolelyforthepersonaluseoftheindividualuserandisnottobedisseminatedbroadly.
306 COHEN, MITCHELL, DOCHERTY, AND HORAN
Warde, 2010). In the present study, greater F0 instability was associated with symptoms that likely reflect increased arousal states in some capacity. While the small correlational magnitudes acrosspatientsmayleadsometodiscountthisfindingforbetweengroups comparison, it is important to note that computerized vocal analysis can be used for monitoring symptoms within patients in a way symptom-ratings cannot. Due, in part, to their near-perfect reliability and high sensitivity, changes in symptom states within individuals can be evaluated using computerized vocal analysis in ways unattainable using Likert-scale based symptom ratings. The modest convergence in symptoms and vocal expression across patients in these measures may belie much stronger convergence within patients, and Pitch Instability may be useful as an index of anxiety, depression, and psychosis fluctuations within an individual—symptoms that are critical to monitor and track. Some limitations of the current project warrant mention. Foremost is our inability to control for individual differences in pharmacotherapy or psychosocial interventions. Notably, there were differences, both across and within studies, in how these interventions were prescribed, how individuals complied with these interventions, and in their effects. While this potential confound reflects noise for the present study, it seems unlikely that treatment variables are responsible for either the null or positive findings in this study. Second, we were unable to disentangle the effects of “context” across the five studies examined here. Variables such as speaking task, research setting, research assistant, compensation, clinical setting, and geographic region may have contributed, or interacted to contribute to the profound differences in vocal expression across studies. Regardless, our conclusions—that vocal expression varies as a function of context (broadly defined), is appropriate. Third, the reliance on the BPRS is a limitation—as it lacks the specificity of other symptom-rating scale based negative symptom measures (Kirkpatrick et al., 2011; Kring et al., 2013). Fourth, the speech topics examined here were relatively neutral in emotional tone, and it is possible that more highly charged topics may have resulted in more pronounced group differences. Finally, the overall sample was relatively low in negative symptom severity, though the general null findings comparing patients with clinically rated negative symptoms versus those without were not dissimilar to the findings from the larger sample. The present findings reflect an important step in applying computerized vocal analytic technologies for understanding and measuring psychiatric symptoms. From a practical perspective, acous
tic analysis is uninvasive, inexpensive, objective, and based on natural behavior that is abundant and closely tied to real-world functioning. Moreover, vocal signal can be procured outside traditional clinical or research office settings using telemedicine and mobile technologies; thus, greatly expanding the scope and potential sensitivity of assessment. The present data suggest that, in contrast to clinician ratings, vocal expression in schizophrenia are largely normal and, at best, occur only in certain contexts. The present data highlight the use of vocal analysis for precisely quantifying negative symptoms and other symptoms as well. At the same time, vocal expression is a dynamic process and what is “abnormal” in one context might be normal in another context. Going forward, it will be critical to consider speaking context when evaluating vocal expression.
References
Allan, E. R., Sison, C. E., Alpert, M., Connolly, B., & Crichton, J. (1998). The relationship between negative symptoms of schizophrenia and extrapyramidal side effects with haloperidol and olanzapine. Psychopharmacology Bulletin, 34, 71–74. Alpert, M., Kotsaftis, A., & Pouget, E. R. (1997). At issue: Speech fluency and schizophrenic negative signs. Schizophrenia Bulletin, 23, 171–177. http://dx.doi.org/10.1093/schbul/23.2.171 Alpert, M., Pouget, E. R., & Silva, R. (1995). Cues to the assessment of affects and moods: Speech fluency and pausing. Psychopharmacology Bulletin, 31, 421–424. Alpert, M., Pouget, E. R., Welkowitz, J., & Cohen, J. (1993). Mapping schizophrenic negative symptoms onto measures of the patient’s speech: Set correlational analysis. Psychiatry Research, 48, 181–190. http://dx .doi.org/10.1016/0165-1781(93)90070-W Alpert, M., Rosen, A., Welkowitz, J., Sobin, C., & Borod, J. C. (1989). Vocal acoustic correlates of flat affect in schizophrenia. Similarity to Parkinson’s disease and right hemisphere disease and contrast with depression. British Journal of Psychiatry, 154, 51–56. Alpert, M., Shaw, R. J., Pouget, E. R., & Lim, K. O. (2002). A comparison of clinical ratings with vocal acoustic measures of flat affect and alogia. Journal of Psychiatric Research, 36, 347–353. http://dx.doi.org/10 .1016/S0022-3956(02)00016-X American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (IVth ed.). Arlington, VA: American Psychiatric Publishing. Andreasen, N. C. (1989). The Scale for the Assessment of Negative Symptoms (SANS): Conceptual and theoretical foundations. British Journal of Psychiatry, 7, 49–58.
Figure 3. Standardized means of vocal variables plotted as a function of patients with No (Non), Mild, and High clinically rated negative symptoms and controls.
ThisdocumentiscopyrightedbytheAmericanPsychologicalAssociationoroneofitsalliedpublishers. Thisarticleisintendedsolelyforthepersonaluseoftheindividualuserandisnottobedisseminatedbroadly.
307VOCAL ANALYSIS AND SCHIZOPHRENIA
Barch, D., & Berenbaum, H. (1994). The relationship between information processing and language production. Journal of Abnormal Psychology, 103, 241–250. http://ccpweb.wustl.edu/pdfs/1994JAbPsych241.pdf. http://dx.doi.org/10.1037/0021-843X.103.2.241 Bellack, A. S., Morrison, R. L., Mueser, K. T., Wade, J. H., & Sayers, S. L. (1990). Role play for assessing the social competence of psychiatric patients. Psychological Assessment: A Journal of Consulting and Clinical Psychology, 2, 248–255. http://dx.doi.org/10.1037/1040-3590.2.3 .248 Boersma, P., & Weenink, D. (Producer). (2013). Praat: Doing phonetics by computer (Version 5.3.59). Retrieved from http://www.praat.org/ Cannizzaro, M., Harel, B., Reilly, N., Chappell, P., & Snyder, P. J. (2004). Voice acoustical measurement of the severity of major depression. Brain and Cognition, 56, 30–35. http://dx.doi.org/10.1016/j.bandc.2004.05 .003 Castelli, F., Frith, C., Happé, F., & Frith, U. (2002). Autism, Asperger syndrome and brain mechanisms for the attribution of mental states to animated shapes. Brain: A Journal of Neurology, 125(Pt 8), 1839–1849. http://dx.doi.org/10.1093/brain/awf189 Cohen, A. S., Alpert, M., Nienow, T. M., Dinzeo, T. J., & Docherty, N. M. (2008). Computerized measurement of negative symptoms in schizophrenia. Journal of Psychiatric Research, 42, 827–836. http://dx.doi .org/10.1016/j.jpsychires.2007.08.008 Cohen, A. S., Dinzeo, T. J., Donovan, N. J., Brown, C. E., & Morrison, S. C. (2015). Vocal acoustic analysis as a biometric indicator of information processing: Implications for neurological and psychiatric disorders. Psychiatry Research, 226, 235–241. Cohen, A. S., & Elvevåg, B. (2014). Automated computerized analysis of speech in psychiatric disorders. Current Opinion in Psychiatry, 27, 203–209. http://dx.doi.org/10.1097/YCO.0000000000000056 Cohen, A. S., Hong, S. L., & Guevara, A. (2010). Understanding emotional expression using prosodic analysis of natural speech: Refining the methodology. Journal of Behavior Therapy and Experimental Psychiatry, 41, 150–157. http://dx.doi.org/10.1016/j.jbtep.2009.11.008 Cohen, A. S., Kim, Y., & Najolia, G. M. (2013). Psychiatric symptom versus neurocognitive correlates of diminished expressivity in schizophrenia and mood disorders. Schizophrenia Research, 146, 249–253. http://dx.doi.org/10.1016/j.schres.2013.02.002 Cohen, A. S., McGovern, J. E., Dinzeo, T. J., & Covington, M. A. (2014). Speech deficits in serious mental illness: A cognitive resource issue? Schizophrenia Research, 160, 173–179. http://dx.doi.org/10.1016/j .schres.2014.10.032 Cohen, A. S., Minor, K. S., Najolia, G. M., & Hong, S. L. (2009). A laboratory-based procedure for measuring emotional expression from natural speech. Behavior Research Methods, 41, 204–212. http://dx.doi .org/10.3758/BRM.41.1.204 Cohen, A. S., Mitchell, K. R., & Elvevåg, B. (2014). What do we really know about blunted vocal affect and alogia? A meta-analysis of objective assessments. Schizophrenia Research, 159, 533–538. http://dx.doi .org/10.1016/j.schres.2014.09.013 Cohen, A. S., Najolia, G. M., Kim, Y., & Dinzeo, T. J. (2012). On the boundaries of blunt affect/alogia across severe mental illness: Implications for Research Domain Criteria. Schizophrenia Research, 140, 41– 45. http://dx.doi.org/10.1016/j.schres.2012.07.001 Cohen, A. S., Renshaw, T. L., Mitchell, K. R., & Kim, Y. (2015). A psychometric investigation of “macroscopic” speech measures for clinical and psychological science. Behavior Research Methods, 1–12. [Advance online publication]. Cohen, A. S., Saperstein, A. M., Gold, J. M., Kirkpatrick, B., Carpenter, W. T., Jr., & Buchanan, R. W. (2007). Neuropsychology of the deficit syndrome: New data and meta-analysis of findings to date. Schizophrenia Bulletin, 33, 1201–1212. http://dx.doi.org/10.1093/schbul/sbl066
Docherty, N. M. (2012). On identifying the processes underlying schizophrenic speech disorder. Schizophrenia Bulletin, 38, 1327–1335. http:// dx.doi.org/10.1093/schbul/sbr048 Dombrowski, M., McCleery, A., Gregory, S. W., Jr., & Docherty, N. M. (2014). Stress reactivity of emotional and verbal speech content in schizophrenia. Journal of Nervous and Mental Disease, 202, 608–612. http://dx.doi.org/10.1097/NMD.0000000000000169 First, M. B., Gibbon, M., Spitzer, R. L., & Williams, J. B. W. (1996). Structured Clinical Interview for DSM–IV Axis I Disorders. Patient Edition. New York, NY: Biometrics Research. Galynker, I. I., Cohen, L. J., & Cai, J. (2000). Negative symptoms in patients with major depressive disorder: A preliminary report. Neuropsychiatry, Neuropsychology, and Behavioral Neurology, 13, 171–176. Giddens, C. L., Barron, K. W., Clark, K. F., & Warde, W. D. (2010). Beta-adrenergic blockade and voice: A double-blind, placebo-controlled trial. Journal of Voice, 24, 477–489. http://dx.doi.org/10.1016/j.jvoice .2008.12.002 Guy, W. (1976). ECDEU Assessment manual for psychopharacology: Publication ADM 76–338. Washington, DC: US Department of Health, Education, and Welfare. Heinrichs, R. W., & Zakzanis, K. K. (1998). Neurocognitive deficit in schizophrenia: A quantitative review of the evidence. Neuropsychology, 12, 426–445. http://dx.doi.org/10.1037/0894-4105.12.3.426 Hoekert, M., Kahn, R. S., Pijnenborg, M., & Aleman, A. (2007). Impaired recognition and expression of emotional prosody in schizophrenia: Review and meta-analysis. Schizophrenia Research, 96, 135–145. http:// dx.doi.org/10.1016/j.schres.2007.07.023 Horan, W. P., Nuechterlein, K. H., Wynn, J. K., Lee, J., Castelli, F., & Green, M. F. (2009). Disturbances in the spontaneous attribution of social meaning in schizophrenia. Psychological Medicine, 39, 635–643. http://dx.doi.org/10.1017/S0033291708003838 Huttunen, K., Keränen, H., Väyrynen, E., Pääkkönen, R., & Leino, T. (2011). Effect of cognitive load on speech prosody in aviation: Evidence from military simulator flights. Applied Ergonomics, 42, 348–357. http://dx.doi.org/10.1016/j.apergo.2010.08.005 Johnstone, T., van Reekum, C. M., Bänziger, T., Hird, K., Kirsner, K., & Scherer, K. R. (2007). The effects of difficulty and gain versus loss on vocal physiology and acoustics. Psychophysiology, 44, 827–837. http:// dx.doi.org/10.1111/j.1469-8986.2007.00552.x Kirkpatrick, B., & Galderisi, S. (2008). Deficit schizophrenia: An update. World Psychiatry; Official Journal of the World Psychiatric Association (WPA),7,143–147.http://www.ncbi.nlm.nih.gov.libezp.lib.lsu.edu/pmc/ articles/PMC2559917/. http://dx.doi.org/10.1002/j.2051-5545.2008 .tb00181.x Kirkpatrick, B., Strauss, G. P., Nguyen, L., Fischer, B. A., Daniel, D. G., Cienfuegos, A., & Marder, S. R. (2011). The brief negative symptom scale: Psychometric properties. Schizophrenia Bulletin, 37, 300–305. http://dx.doi.org/1

…………..Answer Preview………..

……….Answer Preview………….

Why this article

Many are the times parents do not the reasons why their kids experience nonverbal abnormalities. According to this article, these abnormalities are brought about by Schizophrenia as results of studies using scales that rate symptoms have shown that the abnormalities are profound. Therefore, I chose this article to show the actual cause of these abnormalities. This article is very ideal as it uses 5 different studies to arrive at conclusions. With these studies, the article covers all areas about this disease and how those suffering from………..

APA

1,200 Words