Variance In Test Performance Research Articles

NIMG-76. INCORPORATING IMAGING FEATURES OF NORMAL BRAIN TO ENHANCE SPATIALLY DIFFERENTIATING TREATMENT-INDUCED EFFECTS FROM RECURRENT GLIOMA WITH AI

Abstract INTRODUCTION Although AI-based approaches have been applied to discriminate between tumor recurrence (rTumor) and treatment-induced effects (TxE) from chemoradiotherapy in patients with recurrent glioma, these models typically either do not account for within-lesion mixtures of rTumor and TxE or generalize to independent test-sets with >80% accuracy. We hypothesize that incorporating imaging features of normal-appearing brain (NAB) into existing AI-based models that leverage tissue-samples with known locations on MRI and physiological imaging will significantly improve performance for distinguishing regions of TxE from rTumor and allow for visualization of rTumor beyond the T2-lesion. METHODS Using 254 pathology-confirmed tissue-samples with coordinates on diffusion-weighted, perfusion-weighted, and anatomical MRI from 135 glioma suspected of progression, we modified previously-developed binary, AI-based prediction models to learn multi-class targets consisting of rTumor, TxE, and contralateral NAB-regions (3/patient) with imaging signatures of CSF, white-matter, and grey-matter (659 total 10mm-isotropic samples). The multi-class ensemble model was trained with 4 unique repeats of constrained-5-fold-cross-validation and evaluated on a held-out test set (30 patients). This process was repeated 20 times, and the performance was compared to corresponding binary AI-based models. RESULTS NAB was easiest to distinguish from individual and combined classes (AUC-ROCs>0.98). Multi-class model performance for predicting TxE increased by 27% to 0.93±0.03 (p<0.00001) when discriminating between NAB and rTumor as compared to just rTumor with the binary models. While TxE vs rTumor alone was most difficult to predict with the multi-class model (AUC-ROC=0.73) achieving equivalent performance to binary models, there was 67% less variance in test performance. This approach allowed for the quantification of whole-brain probability maps of rTumor, TxE, and NAB along with a combined 3-channel colormap, facilitating the visualization of rTumor predictions beyond the T2-lesion. CONCLUSIONS Including NAB imaging features improved performance of AI-based models of rTumor/TxE, while expanding the predicted area to produce more interpretable whole-brain probability maps. Current work is evaluating their utility in prospectively guiding sampling of rTumor tissue and predicting survival.

The Bilingual Is Not Two Monolinguals of Same Age: Normative Testing Implications for Multilinguals.

A fundamental concept in psychological and intelligence testing involves the assumption of comparability in which performance on a test is compared to a normative standard derived from prior testing on individuals who are comparable to the examinee. When evaluating cognitive abilities, the primary variable used for establishing comparability and, in turn, validity is age, given that intellectual abilities develop largely as a function of general physical growth and neuromaturation. When an individual has been raised only in the language of the test, language development is effectively controlled by age. For example, when measuring vocabulary, a 12-year-old will be compared only to other 12-year-olds, all of whom have been learning the language of the test for approximately 12 years-hence, they remain comparable. The same cannot be said when measuring the same or other abilities in a 12-year-old who has been raised only in a different language or raised partly with a different language and partly with the language of the test. In such cases, a 12-year-old may have been learning the language of the test at some point shortly after birth, or they might have just begun learning the language a week ago. Their respective development in the language of the test thus varies considerably, and it can no longer be assumed that they are comparable in this respect to others simply because they are of the same age. Psychologists noted early on that language differences could affect test performance, but it was viewed mostly as an issue regarding basic comprehension. Early efforts were made to address this issue, which typically involved simplification of the instructions or reliance on mostly nonverbal methods of administration and measurement. Other procedures that followed included working around language via test modifications or alterations (e.g., use of an interpreter), testing in the dominant language, or use of tests translated into other languages. None of these approaches, however, have succeeded in establishing validity and fairness in the testing of multilinguals, primarily because they fail to recognize that language difference is not the same as language development, much like cultural difference is not the same as acquisition of acculturative knowledge. Current research demonstrates that the test performance of multilinguals is moderated primarily by the amount of exposure to and development in the language of the test. Moreover, language development, specifically receptive vocabulary, accounts for more variance in test performance than age or any other variable. There is further evidence that when the influence of differential language development is examined and controlled, historical attributions to race-based performance disappear. Advances in fairness in the testing of multilinguals rest on true peer comparisons that control for differences in language development within and among multilinguals. The BESA and the Ortiz PVAT are the only two examples where norms have been created that control for both age and degree of development in the language(s) of the test. Together, they provide a blueprint for future tests and test construction wherein the creation of true peer norms is possible and, when done correctly, exhibits significant influence in equalizing test performance across diverse groups, irrespective of racial/ethnic background or language development. Current research demonstrates convincingly that with deliberate and careful attention to differences that exist, not only between monolinguals and multilinguals of the same age but also among multilinguals themselves, tests can be developed to support claims of validity and fairness for use with individuals who were in fact not raised exclusively in the language or the culture of the test.

Multiple Sclerosis Performance Test (MSPT): Normative study of 428 healthy participants ages 18 to 89

BackgroundThe Multiple Sclerosis Performance Test (MSPT) is a self-administered, iPad®-based, computerized system for quantifying neuroperformance (cognition, upper and lower extremity motor function, and vision) in patients with multiple sclerosis (MS). ObjectiveThe goal of the study is to provide regression-based norms for the four MSPT test modules to adjust for the influence of demographic variables (age, education, and sex). MethodsThe MSPT was administered to 428 cognitively intact, healthy adults (ages 18 to 89 years). Participants were recruited to achieve a demographically stratified sample from four geographically diverse United States testing sites. ResultsThe amount of shared variance in test performance accounted for by demographic variables was 18-23% for an upper extremity motor test, 31% for a walking speed test, 32% for a low contrast visual acuity test, and 48% for a cognitive test. All four test modules were significantly influenced by age (linear and non-linear effects) and education. Additionally, sex influenced performance on the cognitive and walking speed tests. ConclusionThis study provides regression-based equations that can enhance the clinical interpretation of MSPT scores by adjusting for the potential influences of age, education, and sex.

The role of genetic polymorphisms in executive functioning performance in temporal lobe epilepsy

ObjectiveTo explore the role of several genetic polymorphisms (APOE ε4, BDNF Met, and COMT Val) in executive functioning performance in patients with pharmacoresistant temporal lobe epilepsy (TLE). MethodsNinety-three adults (51 female, mean age = 39 years) with TLE completed executive functioning measures as part of a comprehensive preoperative neuropsychological evaluation, including Trail Making Test (Part B), Wisconsin Card Sorting Test (Conceptual Level Responses and Perseverative Errors), Color Word Interference from the Delis Kaplan Executive Function System, and measures of phonemic and semantic verbal fluency. Genotyping of the APOE, BDNF, and COMT genes was conducted using DNA extracted from peripheral blood or brain tissue (from epilepsy surgery). ResultsAfter adjustment for general cognitive ability, COMT Val carriers showed poorer performance on semantic verbal fluency and color word interference than non-carriers, and BDNF Met carriers showed poorer performance on phonemic verbal fluency than those without a Met allele. SignificanceResults suggest that COMT and BDNF polymorphisms are associated with performance on several EF measures in patients with TLE, including tasks assessing verbal fluency and response inhibition and account for up to 16% of the variance in test performance. The APOE polymorphism was not significantly associated with any of the executive function measures analyzed.

Testing young foreign language learners’ reading comprehension: Exploring the effects of working memory, grade level, and reading task

Extensive research has demonstrated the impact of working memory (WM) on first language (L1) reading comprehension across age groups (Peng et al., 2018), and on foreign language (FL) reading comprehension of adults and older adolescents (Linck et al., 2014). Comparatively little is known about the effect of WM on young FL readers’ comprehension, and even less within testing contexts. Young FL readers are still developing their L1 reading skills and general cognitive skills (e.g., attentional regulation abilities). Completing FL reading tests might be particularly taxing on their WM, and differences in WM capacity – as well as other learner and task characteristics – might create construct-irrelevant variance in test performance.In this study, we investigate the effects of WM, grade level, and reading task on young learners’ FL reading test performances. Ninety-four young English language learners (Grades 6–7) in Hungary completed the TOEFL®Junior™ Comprehensive’s reading test and a WM test battery. Our mixed-effects model predicted significantly higher comprehension accuracy among learners with higher WM capacity, and among learners in Grade 7 compared to learners in Grade 6. Reading task differences were not associated with significant comprehension accuracy differences. We discuss the implications of our findings for testing young learners’ FL reading comprehension.

Factors Associated with Māori Performance on the WAIS-IV.

Neuropsychological tests are routinely used to assess Māori, the indigenous people of New Zealand, yet very few investigations of the psychometric properties of these tests with this population have been conducted. This paper focuses on factors that may impact performance of Māori adults on neuropsychological testing. The Wechsler Adult Intelligence Scale-IV (WAIS-IV) was administered to a sample of 284 Māori stratified for age (between 16 years and 90 years) and gender in order to establish a Māori normative data set. The sample who primarily spoke English as their first language were recruited from a range of urban, regional and rural areas in New Zealand to obtain representation of the main Māori Iwi (tribes). Analysis of test results suggests acculturation, education, income, and, "a culturally positive experience" may have accounted for some of the variance in test performance.

Chapter 6 - Cognitive and neuropsychological examination of the elderly

Neuropsychological assessment plays a prominent role in the evaluation and care of patients with neurodegenerative diseases throughout the dynamic course of disease. As a biomarker of disease, neuropsychological measurement can distinguish normal from pathologic aging processes. Further, neuropsychological data can help distinguish and classify underlying pathologies in dementing diseases, augmenting imaging and biofluid markers in this area. Neuropsychological data can predict increased or reduced risk for dementia conferred by multiple factors, and describe disease trajectory in affected individuals. Cognitive evaluation can also estimate and address functional outcomes that are most important to patients and their loved ones and that are clinically relevant to diagnostic staging. In informing intervention and patient care needs, areas of cognitive weakness highlight targets for support/intervention, while areas of cognitive strength can be capitalized upon to modify the clinical course of disease. These functions can be accomplished through the complementary use of brief screening tools and comprehensive test batteries. However, for neuropsychological data to serve these functions, it is critical to understand neuropsychological test properties and nondisease factors that can account for variance in test performance. This chapter concludes with directions for future research.

Reference values for the incremental shuttle walk test in patients with cardiovascular disease entering exercise-based cardiac rehabilitation

ABSTRACTThe incremental shuttle walk test (ISWT) is used to assess functional capacity of patients entering cardiac rehabilitation. Factors such as age and sex account for a proportion of the variance in test performance in healthy individuals but there are no reference values for patients with cardiovascular disease. The aim of this study was to produce reference values for the ISWT. Participants were n = 548 patients referred to outpatient cardiac rehabilitation who underwent a clinical examination and performed the ISWT. We used regression to identify predictors of performance and produced centile values using the generalised additive model for location, scale and shape model. Men walked significantly further than women (395 ± 165 vs. 269 ± 118 m; t = 9.5, P < 0.001) so data were analysed separately by sex. Age (years) was the strongest predictor of performance in men (β = −5.9; 95% CI: −7.1 to −4.6 m) and women (β = −4.8; 95% CI: −6.3 to 3.3). Centile curves demonstrated a broadly linear decrease in expected ISWT values in males (25–85 years) and a more curvilinear trend in females. Patients entering cardiac rehabilitation present with highly heterogeneous ISWT values. Much of the variance in performance can be explained by patients’ age and sex. Comparing absolute values with age-and sex-specific reference values may aid interpretation of ISWT performance during initial patient assessment at entry to cardiac rehabilitation.

Resilient children are less test anxious and perform better in tests at the end of primary schooling

Resilience has been documented as a key variable in determining successful adaption to adverse circumstances and may also influence successful educational outcomes such as test performance. This paper reports on a study involving 123 pupils from 3 English primary schools, which examined the extent to which test anxiety may mediate the association between resiliency and test performance. Non-verbal ability was measured at the beginning of the school year as a control variable, and test anxiety and resilience via self-report two weeks prior to pupils taking statutory tests in English and maths at the end of primary schooling. Using multiple regression analyses to control for differences in pupils' non-verbal ability, higher resilience was found to predict lower test anxiety and higher test scores. Test anxiety, in turn, was found to predict lower test scores. The lower performance of the less resilient students was mediated by higher test anxiety. These findings indicate how resilience can account for variance in test performance and show how resilience and test anxiety may be related.

Predictors of shuttle walking test performance in patients with cardiovascular disease

ObjectiveThe incremental shuttle walking test (ISWT) is used to estimate cardiorespiratory fitness, but data from healthy individuals suggest that demographic and anthropometric measures account for much of the variance in test performance. The aim of this study was to determine whether anthropometric, demographic and selected gait measures also predict ISWT performance (i.e. distance walked) in patients with cardiovascular disease. DesignObservational study. SettingA community-based cardiac rehabilitation centre (Cohort 1) and a hospital outpatient cardiac rehabilitation programme (Cohort 2). ParticipantsSixteen patients with clinically stable cardiovascular disease (Cohort 1) and 113 patients undergoing cardiac rehabilitation (Cohort 2). InterventionsPatients in Cohort 1 performed the ISWT on two occasions. Anthropometric data and walking and turning variables were collected. Linear regression analyses were used to identify the predictors of test performance. The authors subsequently attempted to validate the equation created by comparing predicted and actual ISWT values in a larger (n=113) validation sample (Cohort 2). Main outcome measuresDistance walked during ISWT, step length and height. ResultsNo gait or turning measures were significantly associated with ISWT performance. Distance walked correlated most strongly with step length (r=0.83, P<0.05) and height (r=0.74, P<0.05). Given the similarity of these correlations and the rarity of step length assessment in clinical practice, ISWT performance was predicted using patient's height; this explained 55% of the variance in ISWT performance. Height was also the best predictor in Cohort 2, explaining 17% of test variance (P<0.01). Body mass index explained an additional 3% of variance (P<0.05) in ISWT performance. ConclusionsRoutine clinical measures, particularly patient's height, are predictive of ISWT performance. The findings of the present study are in partial agreement with similar studies performed in healthy individuals, and it remains unclear whether the ISWT performance of patients with cardiovascular disease is influenced by the same factors as the ISWT performance of healthy individuals.

The cortisol awakening response is related to executive function in older age

Changes in the cortisol awakening response (CAR) have been reportedly associated with older age and may reflect changes in cognitive performance. However methodological issues around adherence, in regard to careful timing of the CAR, suggest caution in drawing firm conclusions. More investigation is also needed regarding which cognitive domains may be most relevant. Executive Function (EF) is linked strongly to functioning of the frontal cortex, itself linked to cortisol secretion via regulation of the Hypothalamic Adrenocortical Axis. In this study, cortisol profiles, cognitive performance and adherence were carefully assessed in a sample of 50 older participants, aged 60–91years (mean=74years). Key aspects of EF were assessed using Form B of the standard Trail-making Test controlling for time taken to complete the simpler Trail-A form of the test. Strong associations between CAR profiles and EF were evident. Cortisol measures significantly predictive of superior EF-related performance in a regression analysis were: earlier peaking and greater magnitude of the CAR. Together these measures explained fully a quarter of all the variance in test performance (R2=0.25; F=7.90; df=2,47; p<.001). Cognitive tests of declarative memory, often linked to hippocampal functioning, were not associated with CAR profiles. We conclude that in a population of healthy older adults aspects of the CAR may be strongly, and perhaps with some degree of specificity, associated with that domain of cognitive functioning, EF, which seems to depend crucially on the integrity of frontal cortex circuitry.

Attention Tests in Different Stimulus Presentation Modes

Tests of attention are frequently used in clinical psychology, neuropsychology, and other applied settings. Depending on the assessment tradition, attention tests are designed and administered differently. Among other aspects, assessment traditions enable testtakers to conduct the tests at either their own pace or at a pace set by the test developer, and apply either rows of stimuli or single stimuli. The current study systematically varied these aspects and examined how these variations affected the structure of attention tests and their correlation with reasoning. A facet model was established that included a facet related to the way attention tests are designed and administered as well as a facet related to the attention ability assessed. Results revealed that both facets explained variance in test performance. Moreover, results revealed that the facet-related test design and administration yielded higher correlations with reasoning than the ability facet. A specific combination of pace and stimulus configuration showed the highest correlations with the applied reasoning tests. Implications for future research and practice are discussed.

Accounting for estimated IQ in neuropsychological test performance with regression-based techniques

Regression-based normative techniques account for variability in test performance associated with multiple predictor variables and generate expected scores based on algebraic equations. Using this approach, we show that estimated IQ, based on oral word reading, accounts for 1-9% of the variability beyond that explained by individual differences in age, sex, race, and years of education for most cognitive measures. These results confirm that adding estimated "premorbid" IQ to demographic predictors in multiple regression models can incrementally improve the accuracy with which regression-based norms (RBNs) benchmark expected neuropsychological test performance in healthy adults. It remains to be seen whether the incremental variance in test performance explained by estimated "premorbid" IQ translates to improved diagnostic accuracy in patient samples. We describe these methods, and illustrate the step-by-step application of RBNs with two cases. We also discuss the rationale, assumptions, and caveats of this approach. More broadly, we note that adjusting test scores for age and other characteristics might actually decrease the accuracy with which test performance predicts absolute criteria, such as the ability to drive or live independently.

How Much do Gastroenterology Fellows Know About Nutrition?

Many people are afflicted with chronic diseases, in which nutrition plays a key role. The need for greater nutrition training among physicians, particularly gastroenterologists, is becoming increasingly evident. To determine the nutritional knowledge and perceived nutrition knowledge of gastroenterology fellows. Thirty-two gastrointestinal (GI) fellows currently enrolled in a GI fellowship program completed a needs assessment evaluating perceived nutrition knowledge and interest in the areas of nutrition support, assessment, obesity, micro/macronutrients, and nutrition in GI diseases. Additionally, an examination evaluating nutrition knowledge specific to gastroenterology fellows was administered. Thirty-two GI fellows completed the needs assessment. Cronbach alpha of the needs assessment instrument was 0.72, indicating satisfactory internal consistency reliability. GI fellows perceived themselves to have the least knowledge in obesity and micro/macronutrients. They indicated a perceived greater knowledge base in nutrition assessment. The mean total test score was 50.04% (SD=7.84%). Fellows had the highest score in the subscale of nutrition assessment (80.64%; SD=19.05%), which was significantly higher than scores obtained in nutrition support (49.45%; SD=11.98%; P<0.05), micro/macronutrients (37.84%; SD=16.94%; P<0.05), obesity (40.11%; SD=20.00%; P<0.05), and nutrition in GI diseases (65.05%; SD=22.09%; P<0.05). A backward linear regression including hours of nutrition education received during GI fellowship, hours of nutrition education received during medical school, and year of GI fellowship accounted for 22.7% of the variance in test performance (multiple R=0.477). Gastroenterology fellows think their knowledge of nutrition is suboptimal; objective evaluation of nutrition knowledge in this cohort confirmed this belief. A formal component of nutrition education could be developed in the context of GI fellowship education and continuing medical education as necessary.

Catechol-O-Methyltransferase Polymorphisms and Some Implications for Cognitive Therapeutics

Catechol-O-methyltransferase (COMT) is a gene involved in the degradation of dopamine and may both increase susceptibility to develop schizophrenia and affect neuronal functions involved in working memory. A common variant of the COMT gene (val 108/158met) has been widely reported to affect prefrontally mediated working memory function, with the high-activity val allele associated with poorest performance across a number of tests sensitive to updating and target detection. Pharmacological manipulations of COMT val 108/158met also have reliably produced alterations in cognitive function, in line with an inverted U function of prefrontal dopamine signaling. Furthermore, there is accumulating evidence that COMT val 108/158met genotype may influence the cognitive response to antipsychotic treatment in schizophrenia patients, with met allele load predicting the greatest improvement with medication. Recently, other single-nucleotide polymorphisms (SNPs) across the COMT gene have emerged as possible risk alleles for schizophrenia, although little is known about whether they affect prefrontal cognition in a manner similar to COMT val 108/158met. Preliminary evidence suggests a modest role for a SNP in the 5′ region of the gene on select tests of attention and target detection. Haplotype effects also may account for a modest percentage of the variance in test performance, and are an important area for future study.

The Paced Auditory Serial Addition Task (PASAT): norms for age, education, and ethnicity.

The Paced Auditory Serial Addition Task (PASAT) is often used to measure attention, concentration, working memory, and speed of information processing. Using a modified 200-item version of the PASAT with presentation rates of 3.0, 2.4, 2.0, and 1.6 items per second, we analyzed demographic influences on test performance in a large sample (N = 566) of healthy North American adults. We found that age, education, and ethnicity were significant predictors, accounting for nearly 23% of the variance in test performance. We discuss these results in comparison to previous normative studies, and present a formula and tables for computing age, education, and ethnicity-adjusted T scores for performance on the PASAT 200.

HIV‐related neuropsychological impairment and immunodeficiency. CD8+ lymphocytes and neopterin are related to HIV‐encephalopathy

The relationship between neuropsychological test performance and immunological parameters was studied in 52 HIV-positive patients within different stages of the infection. All subjects were neuropsychologically tested, the CD4+ and CD8+ lymphocyte count were measured in peripheral blood, and the concentration of neopterin and HIV-p24 antigen were measured in serum. Ten patients with AIDS were defined as neuropsychologically impaired. The CD8+ cell count was the only immunological parameter that could significantly discriminate between AIDS patients with and without neuropsychological impairment. For the total group, significant positive correlations were found between neuropsychological test results and the number of CD8+ and CD4+ lymphocytes, and significant negative correlations were observed between serum concentration of neopterin and neurocognitive function. Regression analyses showed that up to 51% of the variance in test performance could be explained by CD8+ cells and neopterin concentration. The possible role of CD8+ lymphocytes and neopterin in the pathogenesis of HIV-related CNS-dysfunction is discussed.

Self-Report Measures of Ability, Effort, and Nonacademic Activity as Predictors of Introductory Psychology Test Scores

Self-report measures of academic ability, academic effort, and nonacademic activity were used to predict students' performance on their first introductory psychology test. Collectively, these predictor variables explained a significant proportion of the variance in test performance. In addition, academic ability, academic effort, and nonacademic activity each contributed significantly to the prediction of test scores. The relative predictive value of different aspects of academic effort was also examined. Results are discussed in terms of how introductory psychology instructors might advise students who wish to improve their test performance.

Young children's problem-solving strategies: An observational study

The purpose of this study was to identify features of young children's behavior that contribute to effective problem solving. Twenty-four 4- and 5-year-old children, half from low SES families and half from high SES families, were observed while performing perceptual/performance cognitive tasks. Frequency of behavior that reflected problem-solving strategies was recorded. Results are discussed in terms of functional variables that are responsible for SES differences in performance. Problem-solving strategy score was associated with 59% of the variance in test performance. Significant SES differences were found in the strategy score, and when specific behavior categories were examined, SES differences were found in four of the five categories (Visual Scanning, Trial and Error Responding, Impulsive Responding, Helpless Confirmation Seeking).

In Defense of "Private Schools and Public Policy": Reply to Kilgore

Kilgore's critique of our paper (Alexander and Pallas, 1983) raises several questions regarding our conclusion that public schools-do about as well as Catholic schools in promoting cognitive achievement. We agree with Kilgore on only one point: the results we present are not all that dissimilar to those reported by her and her colleagues. What is striking is that we read their implications so differently. Kilgore takes issue with our conclusion that sector differences on test outcomes are small to trivial even before controlling for track enrollment. She is quite right, of course, that the statistics we present are constrained by the variance in the sector measure. We acknowledged our uneasiness with this in a footnote, but nothing in Kilgore's argument moves us from our original position. Consider Kilgore's hypothetical example. She says that the sector variable would account for about 5 percent of the variance in test performance (rather than from .003 to .013 which we report) if there were a sevenfold proportionate increase in the Catholic school enrollment.But who would ever anticipate such a radical transformation of our educational system? And even if -such a remote possibility were granted, is there any reason to expect that achievement processes under that organization would be at all similar to those observed in our data? We concede that our estimations would not apply in Kilgore's example, but that world doesn't exist and probably never will. Our results do characterize the situation that exists at present, though, which we think is far more relevant. Kilgore is not much impressed by our statistics, but she fails to tell us by what standard she judges present-day sector differences to be large. We know of none. Kilgore also criticizes us for not giving her and her colleagues credit for their many attempts to deal with the issue of selection bias. But our critique of their research was quite focused: they failed to deal adequately with the input-level differences in test performance which likely remain after socioeconomic and demographic controls. We do not think any of their procedures are satisfactory on this score, although the analysis they report which controls on curriculum membership (like ours) probably comes closest. Kilgore fails to comment directly on this aspect of our critique. As we said in our paper, she and her colleagues consistently have failed to acknowledge their lack of a suitable input control (or proxy) for competency differences between students in the two sectors. This remains true even through the present exchange and despite the fact that Kilgore's own model in her comment includes predetermined competency levels as an independent predictor of both sector and achievement. We resorted to curriculum controls as an admittedly less than ideal proxy for the elusive datum on input-level test differences. We argued that doing so was preferable to neglecting the issue altogether. Kilgore contends, however, that our procedure fails to credit the Catholic sector with whatever benefits derive from channeling a higher percentage of students into the academic track than is observed in the public sector. This is true and we made note of it in our paper. We argued, though, that there is stronger justification for treating track location as a student choice factor rather than an administrative one if forced to choose between the two. Kilgore offers no reason to think us mistaken. As was stated in our paper, we again grant that school policy differences are relevant to the curriculum placement issue, but we are not much impressed with Kilgore's attempt to take account of this empirically. From a schoollevel analysis, she computes the independentinfluence of sector on curriculum placement and then uses that figure to adjust our results on within-sector achievement differences. Her objective is to derive a measure of sector influence which subsumes both the direct effect of sector on achievement and any indirect effects This research was funded by Grant No. NIE-G80-0113 from the National Institute of Education to the Johns Hopkins University Center for the Social Organization of Schools. The conclusions and opinions expressed are solely the authors' and are not necessarily those of theNIE. Address all correspondence to the authors at the Department of Sociology, The Johns Hopkins University, Baltimore, MD 21218.