Editor’s note: Most Dear Readers of Special Education Today know that I advocate for effective teaching based on solid evidence, not ineffective (or “inadequate”) teaching. Please read this post with that “mind set.” I don’t want y’all to misunderstand what I’m saying. JohnL
On or about 23 September 2025 over on The Learning Dispatch, Carl Hendrick noted that “capable students succeed despite flawed instruction, creating the illusion that those poor methods work, when they only work for those who least need them.” I liked the comment so much that I clicked the “heart” button under it.
Professor Hendrick’s note reminded me of an argument that I sometimes made in classes on learning disabilities when I was a professor in classes for prospective teachers. In this post I recount that argument to make two points: Inadequate Teaching1 (a) explains LD (in part) and (b) creates an illusion of support for itself. There is also a fundamental statistical idea embedded in the post that I want to elucidate.
Why IT persists
Taking point (b) first: we special educators understand why poor pedagogy persists. From a subjective perspective, PP (also known as IT (i.e., “Inadequate Teaching” or “Ineffective Teaching”) looks successful to many educators. It looks like it works because we see so many kids succeed! Here’s an illustration of that subjective view:
Take a group of 25 young children in a classroom employing IT or PP who are being taught, say, spelling according to Method S.2 I’ve divided the students into four clusters here on the basis of what they could do at the beginning of the school year.3
Cluster M includes 4-5 kids who already knew all the fundamental skills of spelling before they walked through the classroom door the first time, so they’ll score high on end-of-year testing. Method S looks like it worked for them
Cluster K includes 8-10 kids walked in the door with all but two or three of the steps on the sequence of skills required to master spelling and will reach Step 10 simply by having more experience with spelling; they’ll get those skills no matter what instruction occurs, and they’ll do well on the end-of-year testing. Method S will seem like it worked for them.
Cluster G includes 4-5 kids who, most of whom had lowish scores on the beginning of the year screener but were showing modest growth on the mid-year assessment. Most of them will discover the fundamental skills and get really close to mastery with Method S because it hints at and exposes them to some of the important fundamental skills over the course of the school year. They somehow (magic?) fill in gaps in their knowledge and sills. Most of them will pass the end-of-year testing. Method S looks like it worked for them.
Cluster D includes 2-3 kids who had low scores on the beginning of the year screening assessment. The teacher “has her eye on” these kids and gives them additional practice pages and other supplements. They will acquire spelling skills for random reasons (they have a sibling who tutors them at home, they repeatedly watch a TV show that demonstrates spelling strategies; one of them might have Special Education Needs and get additional instruction from an SE teacher). Each of them may come close to a passing score or maybe even pass, especially if there is a lot of overlap between the words used by Method S and the words that are used in the year-end test. So Method S can be said to have worked for them.
Cluster B includes 2-3 kids who will struggle all year. They did poorly on the First-month Screener and they’ll grow just a tiny bit by the mid-year progress monitoring assessment. One of them might have previously been identified as having Special Education Needs and got additional instruction from an SE teacher. Because they had abysmal scores on the screener, the teacher will watch them and, when they’re still not taking off on the mid-year progress check, the teacher will give them supplementary practice time with an assistant for “second semester.” They will no not come close to passing the end-of-the year test. So, Method S cannot be said to have worked for them.
So, adding this up, educators can look at the pass rate on their favorite test (is it the NAEP?) and say, “Well, more than 20 of these kids passed! That’s greater than 80% passing! Method S sure looks like it worked for them.”
If we only look at the kids who didn’t begin the school year with “advantages,” though—the 8 students in Clusters B, D, and G—the picture isn’t so rosy. Maybe 3 or 4 of them pass the end-of-the-year test.
OMG, maybe Method S is not so effective!
For those of us with statistical backgrounds, the foregoing illustrates measurement bias. If there is a high probability of an occurrence (“yes, she can read”), then looking only at successes (i.e., occurrences of matches between predicted and actual outcomes) will mislead us. It might just be that the correlation between predicted and observed outcomes is a function of how often those outcomes occur.
When 80% kids learn to read, any method that get's 80% of kids to read will look effective.4
What IT causes
Let’s just guess that Method S is actually not 80% effective. Maybe it’s closer to a coin-toss—50-50 or maybe 60-40. Now thinking about the four or six kids in our Cluster B and Cluster D, we can predict that maybe half of them succeed.
Woohoo. Yay! Great for those 2 or 3 who win the 50-50 coin toss!
What about those 2-3 who didn’t win the coin toss?
Well, let’s just think about them. They must have had horrible home lives…crossed wires…problems getting enough sleep…Oh! Wait. They have LD!5 There is it is. There’s something the matter with those kids.
Of course, it’s really that they have IT.
Summary
Let’s just think through this thought experiment again. Suppose that Method M—an instructional method that has an empirical track record for getting 95% of kids to the status of “reader”—was being used with fidelity in the local education agency.6 Compared to Method S’s track record of, say, 80%, we would expect fewer failures in Clusters K, G, and B. There might still be 1 kid who needs intensive help.
We might call that one kid “LD,” or maybe an instructional failure. And we damn well ought to be providing her or him a special education.
Please understand that I do not consider the number (or percentage) of kids classified as having LD to be a sensible metric for evaluating the quality of instruction. Although it may be tempting to use the illustration I provided here as an indicator that LD is bogus, I’m not making that case. Students are identified as qualifying for special education because of a much more than just their reading performance.
Both the judgment that a student is or is not a reader and the judgement that a student has or does not have LD are statistical statements. As my late colleague, Jim Kauffman and I explained (2024), those judgements depend on establishing a cut point between, let’s say, LD and not-LD. Regardless of where we place that cut point, there will be cases that are just on one side or the other of the line which some advocates would say actually should or should not be said to meet the criterion.
Finally, please understand that what I described here should not be taken as an indictment of teachers. Teachers are not the culprits. Teachers are doing the best that they can. They only have the cards they have been dealt, the initial and on=going professional training they received, the curricula they are authorized to use, the classes of children assigned to them, and the supplemental resources they have available.
What do we educators need? We need secure a better big picture understanding of and support for effective instruction.
Reference
Kauffman, J. M., & Lloyd, J. W. (2024). Statistics, data, and special education decisions: Basic links to realities. In J. M. Kauffman, D. P. Hallahan, & P. C. Pullen (Eds.), Handbook of special education (3rd ed; pp. 81-98). Taylor & Francis.
Footnotes
Some readers may prefer the less fancy term for IT, “Poopy Pedagogy.”
To be sure, the meaning of “taught spelling” has many meanings. We could deconstruct the phrase from now until Ganesha (or later). Let’s just say that “taught” in this context means that the students are receiving some instruction according to some “method.”
The labels for the clusters (e.g., “M”) are meaningless. Please do not assume that they indicate a rank, order, or other evaluative description.
For technically inclined readers, think about the kappa correlation. The key there is that one looks not just at occurrence agreement, but also non-occurrence agreement. Teh app statistic combines both occurrence and non-occurrence agreement so that the resulting statistic is not biased because the event actually occurs often (or not often). We single case researchers among us should be using kappa as measure of interobserver agreement.
Maybe it’s learning differences, not that horrible label, “LD.” Let me move my tongue out of my cheek…sigh.
I am not just making up this argument. There is evidence, as discussed 12 August 2022 on SET, that some educational methods lead to fewer students not meeting criteria for successes than others methods.