Veteran observers of project-based learning (PBL) have anecdotally noted the phenomenon of student "surprises," when teachers are pleasantly surprised that some of their academically-weak students perform well on research projects.

The intent of our pilot study was to better understand the nature of these "surprises." Our guiding questions were, "What characterizes students who perform poorly in conventional classrooms, yet do well in PBL environments?" and "What characterizes teachers who are surprised at the improvement of these students?"

We decided to investigate this phenomenon as a function of the learning styles of teachers and their students (Pask, 1988). We hypothesized that (a) academically low-achieving students who did well in PBL would prefer applied, discovery, technical, or creative (confluent) processing, and that in contrast, (b) their surprised teachers would have learning styles that stressed a preference for acquisition of serialized, factual knowledge.

After three classes of 8th grade science and technology students engaged in PBL work, we asked their three teachers to identify those students who were "surprises" in their PBL performance. They identified 11 such students. Each of these students was interviewed; students as well as teachers were given two learning style inventories to complete. The interviews focused on how the students preferred to learn and included a discussion of what frustrated them in school. The learning style inventories were the 4-MAT (McCarthy, 1980) and the LCI (Learning Combination Inventory, Johnston, 1996).

After correlating the student data with the teacher data, we looked for patterns which would explain the phenomenon under investigation.

The evidence we gathered supported our hypothesis. Students who were identified as "pleasant surprises" were high on the scales for applied, discovery (as measured by 4-MAT), technical, and/or confluent processing (as measured by LCI). These students expressed themselves as follows:

As researchers, we had expected only "pleasant surprises" and were surprised ourselves with an unexpected finding: teachers identified "disappointing surprises" as well. These were academically-successful students who turned in poor projects or no project at all. They scored high on the fact-oriented scale (as measured by 4-MAT). During interviews these students expressed themselves as follows:

Apparently, students who are "pleasant surprises" in PBL have learning style preferences which are not usually allowed expression within the traditional classroom of a teacher who stresses the acquisition of factual knowledge. Their learning styles are given more opportunity to express themselves within the PBL environment. Conversely, it appears that students who are "disappointing surprises" have learning style preferences which find expression within the traditional classroom, but less within the PBL environment. (See patterns #2 and #4 in Table 1.)

Of course, our hypothesis does not explain all of the variance; the human personality is far too complex and dynamic to be categorized neatly. For example, many fact-oriented students do well in PBL environments, and not all students with applied, discovery, technical and/or confluent learning styles do well in PBL. Our pilot study suggests a new perspective on PBL, as well as directions for further research.

?People tend to believe in illusory correlations (Sternberg, 1997, p.83); thus, one may erroneously believe that if a student is a failure in a conventional class, he will necessarily be a failure in PBL. Instead, we suggest that a student may be an unsuccessful learner, not because of a general low ability, but rather because of a learning style which is mismatched with a learning context. We postulate that the one-dimensional belief that "being smart" is context-independent, is the basis for the teachers' surprise.

Our conclusions suggest that teachers should expose students to PBL contexts, thereby: (1) providing a vehicle for previously unsuccessful learners to be successful learners, and (2) stretching supposedly strong learners to develop other important skills. In addition, because teachers in traditional classrooms tend to consistently reinforce a single type of learner, while disadvantaging other types (Sternberg, 1997), teachers should be trained in recognizing and addressing a repertoire of different learning styles.

In order to validate and refine our initial findings, we recommend a large-scale study that incorporates other measures of learning styles, such as "Thinking Styles" (Sternberg, 1997) and "The Five Factor Model of Personality" (Costa and McCrae, 1985).

Costa, P. T., Jr., & McCrae, R.R. (1985). The NEO Personality Inventory Manual.

Odessa, Fl: Psychological Assessment Resources.

Johnston, C., Dainton, G. (1996). The Learning Combination Inventory. Thousand

Oaks, Calif.: Corwin Press, Sage.

McCarthy, B. (1980). The 4-MAT System of Teaching to Learning Styles With Right

and Left Modes. Barrington, Ill.: Excel Inc.

Pask, G. (1988). "Learning Strategies, Teaching Strategies, and Conceptual or

Learning Style" in Schmeck, R.R., Learning Strategies and Learning Styles (pp.

83-100). N.Y.: Plenum Press.

Sternberg, R. (1997). Thinking Styles. New York: Cambridge University Press.

Each of the 6 patterns below (A+B+C) are combinations of student learning styles,

teacher learning styles and learning contexts. These patterns are shown

as predictors of student achievement and teacher reaction. The data of the present pilot study support patterns #2 and #4. The patterns of #1 and #3, while not shown in the study, are assumed to exist. Based on the theory presented in this paper, we would predict that patterns #5 and #6 also exist.

	Independent Variables (Patterns)			Dependent Variables (Predictions)
	A. Student Learning Style Profile	B. Teacher Learning Style Profile	C. Learning Context	Student Achievement	Teacher Reaction
1.	Applied, Discovery, Technical, Confluent	Fact- Oriented	Fact-oriented: Tests	Low	No teacher surprise
2.	Applied, Discovery, Technical, Confluent	Fact- Oriented	Project- Based Learning	High	"Pleasant Surprise"
3.	Fact-Oriented	Fact- Oriented	Fact-oriented; Tests	High	No teacher surprise
4.	Fact-Oriented	Fact- Oriented	Project-Based Learning	Low	"Disappointing Surprise"
5.	Applied, Discovery, Technical, Confluent	Applied, Discovery, Technical, Confluent	Fact-oriented; Tests	Low	No teacher surprise
6.	Applied, Discovery, Technical, Confluent	Applied, Discovery, Technical, Confluent	Project-Based Learning	High	No teacher surprise

Key for Table 1:

Fact-oriented: "What do the experts say?" = Quadrant 2 (4-MAT).

Applied-oriented: "How does this work?" = Quadrant 3 (4-MAT).

Discovery-oriented: "What would happen if?" = Quadrant 4 (4-MAT).

Technical : "Being good with my hands." (LCI)

Confluent: "Being unique and creative." (LCI)