"Intelligence"--that quaint term--has become a source of embarrassment. There is pressure not to acknowledge that variance in human mental ability exists, a feeling that we should not mention such variance, and that if we were really virtuous we would be blind to it.

This is especially true in regard to the group with high intelligence--the gifted. Fashionable attention is aimed the other way, with extensive programs to help those of lesser ability, and these programs have led to a new vocabulary of politically correct speech. In a questionnaire sent to school principals asking how many gifted children they had in their schools, 27 percent said they had none. Because, of course, the next question might be: What are you doing for these gifted students?

The Elephant
It is as if we had an elephant in a classroom, but there were strict rules not to see it. We can imagine the teacher on the first day:

"Now class, it is common that when students first come in, they have the silly impression that there is some large, noisy, lumbering animal in this room. Don't feel embarrassed if you have the same idea [chuckle]. It's only natural.

"But, of course, we all know that there is no such animal outside of old folk tales. You have this idea just because you have been told about it.

"If you're going to do well here, you have to get rid of this idea. Let's not hear any talk about tusks, or trunks, or big gray ears, or the other things that people imagine."

Yet talent searches across the country have made it much more difficult to igore the elephants in our classrooms. Each year, more than 140,000 twelve-year-olds take the SAT or ACT and submit their scores to talent searches and, thereby, make their talents public.

Having found these talented youngsters, what should we do with them? They require special provisions, special classes, and acceleration. Consequently, talent-search programs across the land have begun to offer work tailored for such youngsters.

So now we all recognize the elephant, yes?

No.

There are large numbers of gifted students who require special educational attention. Yet those dictating school policy, and those controlling group-think in education, often reject the very idea of the elephant. This rejection is strident, and it seems to be gaining momentum. Far from supporting the addition of new programs for the talented, the current mood reacts emotionally against the ones we have now.

Tracking
Nowhere is this trend more evident than with regard to the question of tracking or ability grouping: assigning students to different classrooms according to mental ability or subject-matter precocity. In this chapter we will treat the general question of whether students should be instructionally grouped for educational ability; we will not probe here the distinctions some make among grouping, tracking, mainstreaming, and so forth. Our interest is in a broad, psychological treatment of the current debate about instruction and student ability.

This chapter addresses the "detracking" movement in today's schools. As Passow (1988) notes, the literature on this subject ranges "from scholarly reports of research findings to philosophical statements to emotional polemics" (p. 205). Indeed, today the general assault typically emphasizes these polemics, with the charged word "inequality" dominating much of the language of those against ability grouping. Consider an article cowritten by Jeannie Oakes, a professor at the University of California at Los Angeles and perhaps the movement's most visible advocate. This article was published in the Phi Delta Kappan under the title "Detracking Schools: Early Lessons from the Field" (Oakes & Lipton, 1992): "During the past decade, research on tracking and ability grouped class assignments has provided striking evidence that these practices have a negative impact on most children's school opportunities and outcomes. Moreover, the negative consequences of these practices disproportionately affect low-income, African-American, and Latino children" (p. 448; emphasis added). "Striking evidence"? She cites works by others and herself, and two of her own titles give a sense of the movement's intention: Keeping Track: How Schools Structure Inequality (Oakes, 1985) and Multiplying Inequalities: The Effects of Race, Social Class, and Tracking on Opportunities to Learn Math and Science (Oakes, 1990).

Another writer who finds ability grouping culpable is Linda Darling Hammond (1991); "These curricular differences explain much of the disparity between the achievement of white and minority students and between the achievement of higher- and lower-income students.... In this way the uses of tests have impeded rather than supported the pursuit of high and rigorous educational goals for all students." Darling-Hammond cites Lee and Bryk (1988), Oakes, and others to support her remarkable charges, which blame classroom grouping for many of the average differences between race and class groups.

Attackers of ability grouping particularly cite key articles by Robert Slavin (1987,1988,1990a), whose "best-evidence syntheses" are taken as the strongest scientific statements for their side.¹ On the other side, defending ability grouping, are critics of Slavin's methods and conclusions (Gamoran, 1987; Hallina 1990; Hiebert, 1987; Kulik, 1985; Kulik, 1991; Kulik & Kulik, 1982,1984,198 Nevi, 1987; Walberg, 1988).

Although the scientific arguments seem to favor ability grouping, anyone who visits schools sees an apparently headlong movement to abandon it practice. And American citizens, who pay for and suffer from decisions about the schools, decidedly favor the placement of "mentally handicapped" children in special classes. A recent Gallup poll shows 67 percent to 22 percent in support of such tracking (Elam, Rose, & Gallup, 1992, p. 51).

Our profession needs to respond to the claims against tracking, because if they are believed, and class assignments are reorganized without regard to mental ability, the movement away from tracking could greatly affect the future of students, especially of the most gifted students.

Problems of Evidence
The key question is: Does tracking help students or hurt them? Most educators of the gifted would probably say that it helps the abler students. Does tracking hurt other students? Many people claim that it damages poor, black, and Hispanic students, and some would agree with the statement that it destroys "high and rigorous" standards for "all students" (Darling-Hammond, 1991).

Any debate about cause and effect involves problems of evidence. Where should we look for proof? Basically, there appear to be four lines of evidence.

These are:

1.     Indirect argument from principles of measurement

2.     Indirect argument from principles of learning

3.     Direct sampling from experimental evidence

4.     Direct sampling from nonexperimental data sets

Principles of Measurement
Under the principles of measurement, one of the two basic disciplines of educational psychology, ability differences are accepted by virtually all scientists as important and as strongly rooted in biology. These principles have been revealed for generations by Galton, Spearman, Binet, Terman, and many others.²

Almost all scientists agree that intelligence (or g) is easily found as a general or higher-order factor in virtually any set of measures of mental ability, verbal or nonverbal, "fluid" or "crystallized" (Cattell, 1987). G is also widely acknowledged as the best overall predictor of school and occupational success (Ree & Earles, 1992; Schmidt & Hunter, 1992). Indeed, g is well correlated even with measures, such as nerve-response time, that are totally noncultural (see e.g., Jensen, 1992). And the rationale for using multiple competence tests in place of intelligence tests has for the most part been set aside (see, e.g., Barrett & Depinet, 1991).

Opponents of ability grouping commonly claim that mental tests are racially biased, but testing experts are virtually in complete agreement that they are not (see Humphreys, 1992; Linn, 1982). In short, the findings of measurement indirectly support ability grouping in classes as being useful, noninjurious, and nondiscriminatory.

Principles of Learning
In educational psychology, the other major theme is that school learning acts according to certain "laws" common to most learning: principles investigated by Pavlov, Ebbinghaus, Skinner, Thorndike, and many others. Among the widely accepted principles are: working from the simple to the complex, working from the known to the unknown, and starting each lesson near each student's current level of achievement. The idea of individual "readiness" is meant to summarize such principles.

Learning is commonly taken to be individual in locus. It is exhibited in individual behavior and (biochemically) takes place within a single nervous system. Opponents of ability grouping are quite hostile to this emphasis on the individual. One opponent spoke at a session of the American Educational Research Association of the need for "less individualism and more communi-tarianism" (Annual Meeting of the AERA, San Francisco, April 1992). To sidestep this problem of individual differences, opponents of grouping often argue that the best instruction is through "cooperative learning" (see, e.g., Slavin, 1983). In such instruction, the huge problems of heterogeneity are obscured by many student interactions in smaller subsets (each small group having the stronger students teaching the weaker ones). In such classrooms, grades may be given for the performance of the group rather than for the performance of the individual.

Opponents of tracking must reject these established principles of measurement and learning, either by ignoring them or by attacking them (e.g., by claiming that mental tests are biased). But these two major areas of psychological theory, measurement and learning, with their mountains of cumulative evidence and reasoning, provide powerful indirect evidence in favor of grouping students for efficient instruction. And ability grouping has been strongly favored by classroom teachers, surely a sort of testimony that should be properly weighed before drastic changes are made (cf. ERIC Clearinghouse on Tests, Measurement, and Evaluation, 1988, which summarizes teachers' attitudes toward mainstreaming, and Reddick & Pearch, 1984, which reports on nearly a thousand teachers' feelings about tracking).

When researchers study applied questions, however, we commonly wish to go beyond such background principles and to study how various policies have fared in practice. Thus we consider here the two major sources of direct evidence about practices: experimentation in the schools and nonexperimental models using large data sets.

Experimental Evidence
We turn again to one of Julian Stanley's most important contributions to the social sciences: his comprehensive study of research design (see, e.g., Campbell & Stanley, 1963). If we were working in the laboratory sciences, we could perform true experiments, in which the key causal variable (here, ability grouping) would be assigned randomly to the experimental subjects (the students). Then the experimental effect might be estimated from appropriate tests taken by the students.

But in the schools, instruction must take place in classrooms. The appropriate "subjects" would properly be such classrooms, and the evidence would be classroom means on the proper tests. (These are still based on individual scores, and do not contradict the essentially private nature of learning.) In the real world, however, it is unlikely that teachers could be assigned to classrooms at random within a school. Too many other adjustments would be required in the ongoing curriculum.

The assumption behind ability grouping is that the curriculum should be adjusted to the ability levels of the students. If teachers were assigned at random, then it seems unlikely that there would be as much adjustment as when teachers are used to teaching certain defined levels of ability. It seems improbable, then, that the outcome of a random experiment would truly represent the long-term effects of such grouping.

Robert Slavin (1987, 1990b) preferred the data from "experiments," and emphasized those in his "best-evidence" summaries. Thus, for secondary schools (1990b), he reviewed six studies that were "randomized," nine that were "matched," and fourteen correlational studies. His own conclusion was that the results were sufficiently mixed as not to prove overall beneficial effects. He therefore concluded that because his research did not prove benefits in achievement, ability grouping was wrong. In sum, he was opposed to grouping for reasons other than the evidence he collected. He acknowledged, "I am personally opposed to ability grouping, and, particularly in the absence of any evidence of positive effects for anyone, I believe that between-class ability grouping should be greatly reduced" (1990a, p. 506, emphasis added).³

Thus, when Slavin's review failed to convince him of any benefits of grouping, he declared that "there [was] little reason to maintain the practice" (1990b, p. 492). He is often cited by other opponents of grouping, such as Oakes and Darling-Hammond, as if he had indeed proved harm. But this is not the case. Other scholars, frequently studying the same materials as Slavin, have reached quite different conclusions (Feldhusen, 1989, 1991; Gamoran, 1986; Gamoran & Berends, 1987; Kulik & Kulik, 1987).

In any case, the evidence from true experiments is quite thin. Little wonder; such experiments would be disruptive of the usual practices in a school or classroom, and would require much new preparation. Yet paradoxically, an experiment would be invalid if it did not cause school changes from one treatment to another, since it would then not represent the potential differences in curriculum and teaching made possible by grouping.

Nonexperimental Evidence
Because major experimental assignments can be so disruptive, we must look elsewhere for data to test our policies.

The good news is that the federal government has produced massive data sets of highly generalizable materials. These began with Project Talent in the 1960s (the records of which are, however, virtually unreachable by the common researcher). The 1972 National Longitudinal Study (NLS) with five follow-ups through 1986, has been made widely available on computer tapes. And in 1980 there appeared the massive High School and Beyond (HSB), the first in a series of tapes, now out on disks for personal computers. In its follow-ups (1982, 1984, and 1986), HSB also presented much information about school curricula and some aspects of tracking. Still more recent data sets now exist: the National Assessment of Educational Progress (NAEP) and the National Educational Longitudinal Study (NELS, of 1988 and after).

The bad news is this: to study the effects of ability grouping, we ideally need data about both ability and grouping. NLS and HSB are the only data sets with good ability measures--that is, test results largely independent of the achievement measures for the same students. But in these data sets the tracking practices are often difficult to identify. On the other hand, grouping practices are clearer in the more recent data sets (NAEP and NELS), but intelligence measures are nonexistent apart from those for achievement in the various subject-matter areas. In earlier data sets (Talent, NLS, and HSB), student ability was the most useful correlate, the most powerful predictor, for a host of important educational outcomes. It was far more important than income, social class, ethnicity, or school practices.

One fears that ability was too explanatory for political correctness; these nonschool, nonverbal subtests have been removed from the new data sets. Even the brief vocabulary quiz, such a quick and useful measure of general ability, has been stricken from the newer data sets. How unfortunate! And especially unfortunate are the researchers wishing to study ability grouping and its effects on achievement, who have no measures of ability not confounded with the achievement measures they are trying to explain. For this study, we have chosen High School and Beyond, with its good ability measures. (But as noted below, we have had to compromise on the estimation of ability grouping.)

There are other weaknesses in the nonexperimental data with regard to our questions. One is the assignment to classes. Rarely will this be made totally on the basis of an ability score. More often some other factor, such as past accomplishment, will influence a student's assignment to an abler or less able classroom. Such other factors may make the abler classroom appear more effective in its curriculum than it really is, and the weaker class less effective than it is. (This may often lead to a belief that ability grouping had good or bad effects, when both are illusions caused by the methods used to select the high and low groups.)

In short, when we look at the present arguments about ability grouping, we find no persuasive case for those who wish to abolish it. To the contrary, we find that the principles of both measurement and learning argue for ability grouping. And we find no direct evidence, whether experimental or nonexperimental, to justify wiping out ability grouping. Yet there are sources of data for analysis, and we would like, in what follows, to cast more light on such grouping than we have seen in the literature to date.

Ability Grouping or Academic Programs?
We have encountered the same difficulties as others in conducting research on the effects of ability grouping, and, as a result, our evidence about tracking is indirect. But it is also highly relevant to the debate.

The difficulties in conducting research on the effects of ability grouping are magnified at the high school level because students can choose—or are assigned—different types of courses. Thus, the effects of ability grouping are confounded with those of streaming into vocational, general, and academic programs.

But what about such streams? Should they not provide us with some information about ability grouping? Our own research and that of others, for example, suggests that academic coursework is among the most powerful influences on high school students' learning (Keith & Cool, 1992; Keith & Page, 1985).

To narrow the focus: does participation in an academic program, rather than a general-education or vocational program, improve student learning? We attempt to answer that question and then look at its relevance for ability grouping.

Method
For all the analyses discussed here we have used the senior cohort of the 1980 HSB. This data set provides a nationally representative sample of more than fifty-eight thousand high school students, with twenty-eight thousand seniors. For intelligence, this cohort provides fairly robust measures, including some elementary measures little related to the high school curriculum. On the achievement side, however, the measures are of fairly basic skills rather than of the advanced abilities we would prefer to assess. Thus, we have some concerns about ceiling effects for any gifted students in the cohort.

We have conducted ordinary multiple regression analyses, which we will present as path models.

Effects of Academic Programs
Does participation in an academic track improve learning? To answer that question, we simply compared, using a dummy variable, students in an academic track (coded 1) with students in a vocational or general-education track (coded 0).

The model is presented in figure 11.2. We also controlled for students' ethnicity, family background (or socioeconomic status), and intellectual ability. The ability component included two vocabulary tests and two mosaic comparison tests--two nonverbal measures that load as well as the vocabulary tests on a general-intelligence factor. The achievement component included two mathematics tests and one reading test.

The type of program had a substantial effect on student achievement (.22). This is impressive, especially for such a grossly categorized variable as academic program; the .22 figure suggests that one's program in high school has a substantial effect on one's high school academic achievement.

Such programs, of course, are not exactly the same as ability grouping. But what is of interest is the comparison of the effect on students in general with the effect on high-ability students. We next selected only those students who scored a standard deviation or more above the mean. When we re-ran the model for these higher-ability youth, we found what can only be described as a massive .35. Thus, it appears that an academic track is even more beneficial for bright students than for others, and that abler students who are not placed in more challenging courses are being injured in their tested outcomes.

As we noted, this is indirect evidence. But it suggests that opponents of tracking may be off the mark, especially for gifted youth.

Effects of Homogeneous Grouping
The evidence for type of high school program is suggestive, but indirect. We return, then, to the central question: Is clustering by ability good or bad for gifted youth? Another way to look at this question is to ask whether students--especially gifted students--perform better in homogeneous or in heterogeneous ability groups. Refocusing the question as homogeneity versus heterogeneity gives us more flexibility in answering it. HSB does not have adequate data about grouping per se, and thus we cannot resolve this question with our data at the class level. We can, however, broaden the question and examine the effect of homogeneity versus heterogeneity at the school level.

HSB used a two-stage sampling procedure, with schools as the first level, and up to thirty-six seniors selected at random from each school. We calculated the standard deviation of the ability scores for each school and thus created a measure of the heterogeneity of ability at each of the nearly one thousand schools in HSB. Then we reversed the standard deviation of ability so that homogeneous schools received high scores and heterogeneous schools received low ones. We then used this new variable to examine the effect on student achievement of the homogeneity in ability--the underlying theme of ability grouping.

Effects on All Students
The first analysis held no surprises. Homogeneity in ability had a positive effect on achievement (see figure 11.3), but it was a small one. Indeed, even though this effect is statistically significant, we generally do not consider paths below .05 meaningful. This finding corresponds quite well with that of Kulik and Kulik (1987): ability grouping shows a (rather slight) overall favorable effect on achievement.

High-Ability Youth
But our concern in this research is primarily with gifted youth rather than youth in general. Is homogeneous grouping helpful or harmful for brighter students? When we selected only the top 16 percent of seniors, our analyses painted a quite different picture. When we examined the effects of homogeneity on high-ability students, we found a moderate influence of homogeneity on achievement (.13; see figure 11.4). (This and subsequent figures focus only on the effects of variability in ability; other effects are not included.)

Greater variability produces lower achievement, and greater homogeneity produces higher achievement. That is, high-ability students perform better when they are in a homogeneous, rather than a heterogeneous, environment.

Opponents of ability grouping may argue that these findings are well and good, but that ability grouping is harmful for minority students. When we examined the effect of homogeneity on high-ability black youth, however, we found it had a much stronger effect on these students than on high-ability students in general. Whereas homogeneity has a moderate positive effect on all high-ability youth (.13), it has a very strong positive effect on high-ability black youth (.32; see figure 11.4). This powerful effect suggests that we should oppose heterogeneity and support grouping. Also, we found a substantial effect in favor of grouping for high-ability Hispanic youth (.24; see figure 11.4).

"Well," our hypothetical opponent of grouping might say, "grouping may be good for high-ability youth, but it undoubtedly is harmful for low-ability youth." Again our data disagree. Ability grouping had no substantive positive or negative effect on low-ability students in general (.00), on low-ability black students (.01), or on low-ability Hispanic students (.01), Contrary to current conventional wisdom, surrounding a low-ability student with a homogeneous group of students seems to have no effect. Homogeneous grouping is not apparently helpful for such students, but neither is it harmful.

Other Outcomes
Our opponent might argue (as opponents have time and again) that our analyses are too narrow because they focus only on achievement as an outcome. Does research not show that grouping is detrimental to students' self-esteem and aspirations? Such arguments rest on assumptions about the ill effects of grouping or of other systems of classification or labeling. For a deep analysis of the defects of "labeling theory," see Gordon's (1980) critique. But what of our own findings with the HSB data? Again, we could find no support for such arguments.

Educational Aspirations
We studied the effects of homogeneity in ability on students' educational aspirations: how far they and their parents want them to go in school (see figure 11.5). On students in general there were no effects of homogeneity in ability on aspirations. On high-ability students there was a small positive effect, and on high-ability Hispanic students there was a moderate positive effect. The effect on high-ability black students was insignificant.

Self-Concept
We also studied the effects of homogeneity on students based on a four-item self-concept scale in HSB (cf. Pottebaum, Keith, & Ehly, 1986). Grouping had no significant effect on students in general, on high-ability students, on high-ability black students, or on high-ability Hispanic students (see figure 11.6).

Locus of Control
Similarly, we studied the effects of homogeneity on students based on a four-item locus-of-control scale. On students in general and on high-ability students, being in a homogeneous group had a significant, but nonmeaningful, positive effect on locus of control. Homogeneous grouping had insignificant effects on the locus of control of high-ability black and Hispanic students.

In summary, we could find no evidence that grouping--attending schools with a homogeneous instead of a heterogeneous group of students, as ascertained by our ability measures--had any negative effect on these important nonacademic criteria. Indeed, the only significant effects we found are in favor of such grouping. Although supporting data are not presented here, there were also essentially no effects on students of low ability.

Caveats
We have already noted the problems and dangers of nonexperimental research, and we should, therefore, be cautious in our interpretation of these findings. Critics of grouping might argue that our research is flawed in other ways. Perhaps we find these effects because our ability measure is verbally weighted. Our hypothetical critic might argue that we should instead examine nonverbal ability, because it may be a "fairer" yardstick for minority youth. In fact, the effects in favor of grouping were even stronger when our examinations used the control of nonverbal ability (mosaic comparisons) alone.

Perhaps homogeneity of ability is simply a proxy for the average ability of the school. That is, is it not simply high-ability youth going to school with other high-ability youth that makes the difference? No. The addition of the mean ability of the school in our models made no substantive difference in these results.

Critics can point out that it is not really grouping we are studying. Our analyses have concerned the homogeneity of ability at the school level rather than at the classroom level. Our choice to study homogeneity was one of necessity: HSB simply had no adequate measures of grouping, unconfounded with other variables such as students' high school streams. And in some ways, homogeneity may be a preferred variable to study. It is, after all, at the heart of the controversy: Do students learn better in an environment with students similar to themselves, or with those different from themselves?

In addition, if we had been able to control for grouping at the classroom level rather than at the school level, our results--generally in favor of grouping for high-ability youth--would logically have been stronger rather than weaker. Undoubtedly some of the highly heterogeneous schools practice grouping at the classroom level, thus gaining some of the positive effects of homogeneity. Therefore, the positive effects found here may underestimate the true effects of grouping.

Finally, these results in favor of homogeneous schools were obtained with fairly basic measures of reading and math achievement. If the HSB tests had contained measures of advanced knowledge in these and other subject matters, our effects would again likely be stronger. Better measurement of grouping and achievement would likely result in stronger, not weaker, estimates of the positive effects of homogeneous ability groups on high-ability youth.

Summary
Schooling in a homogeneous group of students appears to have a positive effect on high-ability students' achievements, and even stronger effects on the achievements of high-ability minority youth. Grouping does not seem to affect negatively the achievements of low-ability youth. Indeed, ability grouping seems to have no consistent negative effects on any group or any outcome we studied.

Therefore, we reject the claims of opponents of ability grouping that it is harmful to students' achievements, aspirations, or self-perceptions. Instead, we assert that ability grouping may have positive effects on gifted students' learning, the most important educational outcome, and that these effects seem particularly powerful on gifted minority youth. If grouping indeed has positive effects on high-ability youth and no negative effects on low-ability youth, we can see no reason to support the current trend away from ability grouping. Talents are far too rare, and too valuable for society, to be sacrificed on an altar of blind egalitarianism.

Notes

1.     But see Slavin, Madden, Karweit, Livermon, & Dolan (1990) for a remarkable contradiction of the antitracking principle: in the "Success for All" program, which he reports as highly effective, primary students are put into reading groups for ninety minutes a day on the basis of reading level. Especially note the reasons Slavin and his colleagues claim for its success: "The idea behind regrouping is to allow teachers to teach the whole reading class without having to break the class into reading groups. This greatly reduces the time needed for seatwork and increases direct instruction time. ... It does ensure that every reading class will be at only one reacting level, eliminating workbooks, dittos, or other follow-up activities that are needed in classes with multiple reading groups" (emphasis added).

Slavin's own strategy in practice is especially notable in sweeping aside any social considerations: "a reading class might contain first-, second-, and third-grade students all reading at the same level" (p. 259, emphasis added). The physical and social differences between first and third grade are, of course, tremendous. And both children and their families will be well aware of the dramatic group differences.

2.     This is not the place to argue these elementary and other principles, since they may be found in most advanced texts on measurement, learning, behavior genetics, and similar subjects.

3.     As the scholar can easily note, there often is a division along disciplinary lines, with sociologists usually opposed to grouping (see also Kerckhoff, 1986), and measurement specialists more often in favor of it. But there are distinguished exceptions on both sides.

*Please see original chapter for all figures.

Acknowledgments
Thanks are given to Robert Gordon, John S. Lutz, Harry Passow, and Betsy Becker for useful comments and information.