Neurological Basis of Deficits in Affective Prosody Comprehension Among Alcoholics and Fetal Alcohol–Exposed Adults

Subjects were recruited from two Veterans Affairs Medical Center facilities and were either patients or employees. Subjects agreed to participate in an approved study of emotional expression in speech and signed an informed consent form that was approved by the Institutional Review Board of the University of Oklahoma Health Sciences Center and the Veterans Affairs Medical Center in Oklahoma City. After screening for neurological injury or illness, subjects were scheduled for the 1.5-hour testing session. Participation was limited to individuals under age 65 years because deficits in APC have been discovered in the normal elderly population.⁵² Subjects were 25 to 63 years old, were either African American or Caucasian, and were all right-handed. Detoxified alcoholics (ALC) with 21 days or more of sobriety who met criteria for a DSM-IV alcohol abuse diagnosis⁵³ were scheduled for testing. Subjects who reported information indicating their mothers had abused alcohol during pregnancy were placed in a group of fetal alcohol exposure subjects (FAexp), whether or not they themselves had a history of alcohol abuse. The criteria for FAexp group placement included family history information such as mother's incarceration for public drunkenness and loss of custody of minor children due to alcohol abuse. Subjects with a Quantity Frequency Index of less than 5 ounces of absolute ethanol per drinking day were placed in the control group (CON) if they reported no history of alcoholism treatment or fetal alcohol exposure. Characteristics of control and alcohol-exposed subjects are shown in Table 1. In order to provide comparisons with other populations, 10 left hemisphere brain-damaged patients (LBD) and 9 right hemisphere brain-damaged patients (RBD), age-matched to the ALC and FAexp groups, were included in statistical analyses of subtest performance patterns. These patients were selected from previously documented studies.^20,54

Exclusion criteria for ALC and FAexp groups included serious neurological injury. In addition, all five subject groups were screened for neurological illness, chronic illnesses that result in significant restriction of oxygen to the brain, and DSM-IV psychiatric disorders other than substance abuse, and subjects were excluded if any of these were present. Medical records were examined for evidence of neurological injuries or serious illnesses; one alcoholic was removed from the final study group because of a closed head injury.

A health questionnaire was used during telephone screening for inclusion/exclusion criteria. The Shipley Institute of Living Scale⁵⁵ was used to screen subjects for cognitive impairment. The Alcohol and Substance Use Questionnaire, a structured interview instrument (copyrighted by the Oklahoma Center for Alcohol and Drug-Related Studies at the University of Oklahoma Health Sciences Center), determined extent of substance abuse in volunteers and their extended families. Thirty-two alcoholics (ALC) reported data indicating that their mothers did not abuse alcohol at any time during early child-bearing years. However, 11 subjects reported concrete behavioral data indicating that they were fetally exposed to alcohol (FAexp). Of these subjects, 9 were recovering alcoholics and the other 2 reported no lifetime alcohol use disorder. Forty-one control subjects (CON) were included in the final project group.

The Aprosodia Battery²⁰ assesses production, comprehension, and discrimination of affective prosody. The two major sections of the Aprosodia Battery are Production (spontaneous production and repetition of affective prosody) and Comprehension (identification and discrimination of conveyed emotion in speech). Subjects listen to compact disk recordings of three sets of utterances. The first Comprehension subtest, Word, uses a completely articulated sentence, “I am going to the other movies,” that is repeated using random variations of six emotions. The six emotions are Happy, Sad, Angry, Surprised, Disinterested or Bored, and Neutral. In addition, two stress patterns, with emphasis on “am” and with emphasis on “other,” test the subject's ability to suppress attention to linguistic prosody cues. Each exemplar is presented twice, resulting in 24 sentences in each subtest. Subjects indicate which one of six emotions they hear; exemplars can be repeated so that attentional variables are reduced. Twenty-four monosyllabic (“ba ba ba ba ba“) and 24 asyllabic (“aaaaahhhhh”) utterances with the same emotional and stress patterns are also presented to the subject. The fourth subtest, Discrimination, presents 24 pairs of low-pass filtered exemplars from the Word subtest. The 12 affective-stress sentence combinations comprising the Word Comprehension stimulus set were played through a 70–300 Hz bandpass-filter and recorded. This procedure preserves prosodic-acoustic information involving intonation and intensity at both global (affective) and local (stress) levels while markedly reducing phonetic information. A stimulus compact disk was then made comprising 24 randomized sentences: 12 with the same affective intonation but different stress patterns and 12 with different intonations but the same stress pattern. For each of the 24 pairs, subjects must tell if the two sentences use the same or different emotional categories. Because these sentences employ different stress patterns, the task also tests the ability to discriminate between affective and linguistic prosody.

The number of correct responses on each of the four Affective Comprehension subtests of the Aprosodia Battery (Word, Monosyllabic, Asyllabic, and Discrimination) were the primary dependent variables used in this study. Z-scores were computed based on the CON group mean and standard deviation. Both SAS⁵⁶ and SPSS 8.0 for Windows (SPSS, Inc.) statistical programs were used to analyze data. For the variables reported in Table 1, a mean was calculated for normal distributions, and a median to more accurately describe sample characteristics when outliers distorted the distribution. Variables were assessed by skewness figures⁵⁷ and the Shapiro-Wilk W Normality Test⁵⁶ for their appropriateness in parametric analyses. Nonparametric tests used when variables did not show normal distribution were the Wilcoxon/Kruskal-Wallis tests (rank sums), median test (number of points above median), and Van der Waerden test (normal quantiles). Variance equality was assessed via the Bartlett; the conservative Welch analysis of variance (ANOVA) was used to test means when the variances were unequal. Post hoc analyses of estimated marginal means using Student-Newman-Keuls (homogeneous subsets and harmonic mean sample size with alpha=0.05) were used to ascertain significant differences on subtest scores for each group. Means comparisons were also assessed by using the Tukey-Kramer honestly significant difference test (significance level at <0.05) and the Student-Newman-Keuls post hoc analysis.

Group characteristics are shown in Table 1. Gender, ethnicity, and years of education were not associated with Aprosodia Battery scores among control subjects; one-way ANOVA and regression analyses performed on each of these variables with the four Comprehension subtests resulted in P>0.10 on all tasks.

The scores earned by the three groups (CON, ALC, and FAexp) on the Shipley Institute of Living Scales (Table 1) were analyzed by use of one-way ANOVAs. No significant differences were found between the groups on vocabulary, abstract reasoning, and a measure of cognitive impairment.

The research goal was to determine if the profiles of scores from the four Aprosodia Battery Comprehension subtests for ALC and FAexp groups were similar to an LBD/Callosal profile, an RBD/Aprosodia profile, a combination of the two, or a unique profile. To explore this question, the patterns of deficits of ALC and FAexp subjects were compared with data previously collected on patients with unilateral ischemic strokes who were tested on the Aprosodia Battery at 4 to 6 weeks post infarction.^20,54 Because the research interest was on subjects with abnormal performance, the experimental groups were limited to individuals who scored <–1.64 SD (normal curve proportion <0.05) on at least one of the four Comprehension subtests. Group sizes, subtest means, and profiles are displayed in Figure 1. The data were subjected to a 5×4 repeated-measures ANOVA to compare means. A significant task by group interaction was observed (F=5.60, df=12,264, P=0.000), in addition to significant main effects for group (F=53.98, df=4,88, P=0.000) and task (F=32.03, df=3,264, P<0.000). Post hoc analyses using multiple 2×4 repeated-measures ANOVAs assessed how each patient group performed compared with CON. These tests showed highly significant task by group interaction (all F>8.45, all P<0.001), and main effects for group (all F>61.62, all P<0.001) and task (all F>8.49, all P<0.001) for the ALC, FAexp, and LBD groups. For RBD, however, no significant task by group interaction (F=0.45, df=3,144, P=0.71) or main effect for task (F=0.46, df=3,144, P=0.71) was observed, but a highly significant main effect for group was found (F=99.89, df=1,48, P=0.000). To assess how the four patient groups performed against each other, a 4×4 repeated-measures ANOVA was performed to compare means. It showed a significant task by group interaction (F=2.54, df=9,144, P<0.01) and highly significant main effects for both group (F=10.58, df=3,48, P=0.000) and task (F=19.03, df=3,144, P=0.000). A Student-Newman-Keuls post hoc analysis (alpha=0.05) was performed, with results shown at the bottom of Figure 1.

Although the post hoc findings are somewhat complex, the following is evident by examining Figure 1. The ALC profile (open squares) is not significantly different from the LBD profile (closed squares) for the Monosyllabic, Asyllabic, and Discrimination subtests, but shows a significant divergence on the Word subtest that suggests a partial RBD component. The FAexp profile (open triangles) is not significantly different from the RBD profile (closed triangles) for the Word, Monosyllabic, and Asyllabic subtests, but shows divergence on the Discrimination subtest that suggests a partial LBD component. Therefore, subtest performance on the Aprosodia Battery by ALC and FAexp could represent a combination of right and left brain-damaged profiles.

To determine if this supposition was correct, the LBD and RBD data were collapsed into a single brain-damaged group (BD). The subtest means and profiles are displayed in Figure 2. A repeated-measures ANOVA was performed comparing BD, ALC, and FAexp group means. This time, the results showed no significant task by group interaction (F=0.41, df=2,49, P<0.67). However, significant main effects for group (F=15.07, df=2,49, P<0.000) and task (F=42.61, df=2,49, P<0.000) were still present. Since no interaction was observed, the Comprehension profile patterns of BD, ALC, and FAexp were statistically identical, supporting the supposition that alcohol-exposed subjects have a combination of LBD/Callosal and RBD/Aprosodic deficits.

The causes for the significant main effects were assessed by collapsing the data two ways—across groups preserving tasks and across tasks preserving groups—followed by separate one-way ANOVAs and Student-Newman-Keuls post hoc analyses (alpha=0.05) because no task by group interaction was present. The main effect for group was due to less severe overall deficits in ALC (mean Z-score=–1.52) compared with FAexp (mean Z-score=–3.12) and BD (mean Z-score=–2.82), which did not differ from each other. The main effect for task was due to the similar performance on the first two subtests, Word (mean Z-score=–2.94) and Monosyllabic (mean Z-score=–2.83), both significantly impaired compared with Asyllabic (mean Z-score=–2.15), which in turn was statistically worse than Discrimination (mean Z-score=–1.11).