I have a few comments on Benjamin Barton’s conclusion that there is no correlation between prolific writing and effective teaching.
First, Barton uses student ratings as the only measure of teaching effectiveness. He does not establish the degree to which student ratings are useful measures of teaching effectiveness, if effectiveness means that students learn facts and processes of thinking that are more important to learn than other facts and processes they might learn with equal investment. The connection has employed educators for decades and the correlation between student ratings and teaching quality is too weak, especially across various methods of teaching and types of classes, to justify using ratings as the sole basis for policy.
Second, student ratings may be biased. Suppose that students shift their evaluations upward for teachers who write more, as well as for teachers who teach better. A teacher could shift time away from teaching to writing, reducing the quality of teaching, without that reduction in quality being reflected in his ratings. Or, just to take the other side, suppose that people who write more are more aware of uncertainty and present ideas less definitively, lest the students get the wrong idea. As such a teacher writes more, her actual teaching quality rises, but her increasing indeterminacy prevents her ratings from rising. Or suppose that teachers who write more teach worse but teach more upper level elective courses which students tend to rate more highly. In all three examples, the true correlation is hidden.
Third, suppose that for teachers who publish little, writing more would improve teaching, while for teachers who publish much, writing less would improve teaching. Aggregating the groups would make it look like writing more had no effect on teaching, even though it had an effect within both subgroups. Lindgren’s focus on subpopulations might better reveal actual relationships within those subgroups, although such results should not be generalized to the whole population of teachers.
Fourth, a meaningful relationship between teaching and writing within a small subgroup might be masked by all of the data in the rest of the spectrum. For those who write a lot, shifting even more time to writing might decrease teaching quality, yet that effect might be masked by the large group of teachers who write little and for whom increasing writing does not detract from teaching. For some faculty members, increasing publishing does not mean decreasing teaching preparation. For others it does. If the former group far outnumbers the latter, the lack of a relationship within the former may mask the relationship that does exist in the latter group.
What almost *has* to be true? If a professor is working the maximum number of hours already, spending more time on teaching means less time on writing. How many teachers working to the limit have found that teaching is best when NO time is spent preparing for class? Starting from zero class preparation time, teaching will be improved more by spending five minutes on class than by spending that same time writing. Some class preparation is better than none.
However, by encouraging rigor and allowing analysis to be carried beyond what can be held in memory, writing often teaches a writer lessons he would not have learned had he not committed ideas to paper. It seems likely that many teachers will teach better when they spend at least a little time writing for publication. Even when only teaching quality matters, some writing beats none.
These points might form the foundation of a better model. There is a non-zero optimum level of research, and moving away from it will reduce teaching quality. To test it, we might look at those who write the most and try to see whether writing even more reduces student ratings. Looking at those who write least, we might try to determine whether writing more improves student ratings. Increased writing may have different effects on teaching at the two ends of the spectrum, and aggregating all teachers could wipe out evidence of those effects. In short, the finding of no correlation does not eliminate the possibility of a complex causal connection.
Jeffrey E. Stake