January 8, 2009

Spelling tackled in American Educator

Following my lead, American Educator has a good article on Spelling and how to teach it.

Not unsurprisingly, popular spelling instruction practices are based on flimsy pseudo-science:

One common perception we have encountered is that visual memory, analogous to taking a mental picture of the word, is the basis of spelling skill. Teachers often tell us that they teach spelling by encouraging whole-word memorization (e.g., using flashcards and having students write words 5 or 10 times) or by asking students to close their eyes and imagine words. We’ve encountered this perception that spelling relies on visual memory so many times that we became curious about when and how it originated—after all, it’s a far cry from Webster’s spellers. We traced it back to the 1920s: one of the earliest studies to stress the role of visual memory in spelling was published in 1926, and it found that deaf children spelled relatively well compared with normal children of similar reading experience.4 Based on this study, and the perception that the relationship between sounds and the letters that spell them is highly variable, many people concluded that learning to spell is essentially a matter of rote memorization. Thus, researchers recommended that spelling instruction emphasize the development of visual memory for whole words


Right, let's have them visually memorize whole words because there couldn't possibly be any other helpful information they could use. Apparently not; its a common belief that English is a highly irregular language. The article lays that trope to rest.

This is a question we hear often. If English spelling were completely arbitrary, one could argue that visual memorization would be the only option. However, spelling is not arbitrary. Researchers have estimated that the spellings of nearly 50 percent of English words are predictable based on sound-letter correspondences that can be taught (e.g., the spellings of the /k/ sound in back, cook, and tract are predictable to those who have learned the rules). And another 34 percent of words are predictable except for one sound (e.g., knit, boat, and two). If other information such as word origin and word meaning are considered, only 4 percent of English words are truly irregular and, as a result, may have to be learned visually (e.g., by using flashcards or by writing the words many times).

Far from being irregular and illogical, to the well-known linguists Noam Chomsky and Morris Halle, English is a “near optimal system for lexical representation.

...

There are three types of information that, once learned, make spelling much more predictable: (1) word origin and history, (2) syllable patterns and meaningful parts of words, and (3) letter patterns.”


It doesn't take a rocket scientist to figure this out. Research shows that children misspell irregular words more often than regular words. That should have been a good indication that visual memorization might not have been the best way to go.

The other thing is. Wouldn't using word origin and history, spelling patterns and meaningful parts of words, and letter patterns to spell words involve using and practicing critical thinking skills -- dare I say 21st Century skills, rather than brute memorization? Just sayin'.

The article is a good read. The only weak part is when the authors make some untested recommendations as to how they think spelling should be taught. At best, these recommendations are representative examples of what might possible be good practice once someone takes to time to develop an test a suitable instructional sequence. But, that work has not yet been done and the authors are a wee bit overconfident that their recommendations will be effective.

15 comments:

Stephen Downes said...

This is the significant bit:

> There are three types of information that, once learned, make spelling much more predictable: (1) word origin and history, (2) syllable patterns and meaningful parts of words, and (3) letter patterns.”

The commentary builds on this:

> Wouldn't using word origin and history, spelling patterns and meaningful parts of words, and letter patterns to spell words involve using and practicing critical thinking skills -- dare I say 21st Century skills, rather than brute memorization?

Short answer: yes.

Words have structure and (internal) syntax. They are built through common mechanisms. The spelling of words, in other words, can be understood by a process of recognizing patterns in words.

Simple memorization of word forms actually makes it much more difficult to remember how words are spelled, because there is no means of finding anything in common from one word to the next. To a certain (lesser) degree, simple memorization of word sounds fares poorly as well. The patterns in word spelling have to do with - as stated above - origins, meaningful parts of words, spelling patterns, and more.

To take this discussion a step further, to explain why the connection between this and "21st century knowledge"...

If we were trying to understand spelling the 'old' way, then once we had understood that spelling can be informed by looking at eytemology, functions of parts of words, etc, we would have then attempted to create a set of 'rules' to teach children to spell (the way we did, for example, for grammar).

Learning to spell well would then be a matter of learning the rules, then applying them appropriately.

The '21st century approach', however, does *not* view spelling as rule-based. That does not mean that we cannot identify commonalities - we can and we should. Rather, the 'rules' are descriptive rather than prescriptive. They inform us about the nature of the word, rather than telling us how to spell the word.

The teaching of spelling thus becomes a matter of teaching pattern recognition, of clustering, rather than the teaching of rules. It is a process of presenting collections of words in which students can identify similarities, and presenting a diverse selection of words so students can recognize varying clusters of similarities.

Teaching thus allows students to *continue* to learn how to spell words as their vocabulary increases, to infer to patterns of spelling forms from their own experience.

KDeRosa said...

It is a process of presenting collections of words in which students can identify similarities, and presenting a diverse selection of words so students can recognize varying clusters of similarities.

Sounds like you agree with Engelmann's theory of instruction.

Am I right?

Anonymous said...

It's a bit presumptuous to term this "Englemann's Theory," Ken. Zig has made important contributions, but he, Carnine, and the other Oregonians have added little to the theory as expounded by BF Skinner. And Skinner was always careful to recognize the contributions of forerunners that trace back to E. L. Thorndike.

One may not "like" the theory. But it's scientific basis is unassailable, and "cognitive psychology" and "neurobiology" have done nothing to add positively to either instructional theory or practice in the interim.

The difficulty comes in moving from the theory into practice, as is true with any theory. Zig's programs are one way to do this, but they are not by any means the only way.

As I've said, you have to go to the UK to find others these days, and even there you have to look hard. The professional and public politics mitigating against reliable instruction are strong.

Meanwhile, back at the spelling ranch. It's the structure and substance of the Alphabetic Code that controls English spelling.

Word origin and history is relevant, but it won't teach spelling.

English is alphabetic based rather than syllabic based. Morphemic characteristics are important, but "syllabic patterns and meaningful parts" only murk up the instruction.

It's not "letter patterns" by phoneme/grapheme correspondences in spelling and grapheme/phoneme correspondence in reading that are relevant.

KDeRosa said...

Dick, I was under the impression that Zig's contribution was the faultless instruction/logical analysis part which was the part, along with Becker's praise component, that got Skinner's behavioralism to work -- something Skinner was unable to do. As such, I'd say that was a major contribution and something apart from what Skinner was doing -- not just a practical application. Though, I agree that Skinner and others laid the foundation.

Correct me if I'm wrong. (Like I have to ask.)

Libby Maxim said...

hey, why not ask the kids who win the spelling bee what they do to spell so well,

seems to me that observing these kids in action and practice, might teach us a few things

libby

Anonymous said...

"faultless instruction/logical analysis" was a 1963 contribution of Herb Terrace, a Skinner student.

The flaw is that instruction is not dealing with a constant like one concept at a time. The expertise morphs as it goes along. Holding the learner's hand too tightly generates the kinds of scripts than most teachers find deadly.

In learning/teaching any complex skill some errors as the complexity of the task increases are desirable. The student is learning to handle the variations of the environmental context.

Zig's "big mistake" was in going along with standardized achievement tests and relative comparisons as the gauge of instructional accomplishments.

"Praise" goes way back to E. L. Thorndike. Becker had some useful insights on how and when to dispense praise to control behavior.

Becker's "big mistake" was dying at an early age.

Anonymous said...

Libby says, "Ask spelling bee winners."

Good idea. You can find out about almost anything on the Internet. As with all outliers, it turns out that "determined practice" is key.

Here are 2 quotes from winners:

“To prepare for the spelling bee, he said he stayed up late studying for days.”

“The amount of studying, memorization and love of the English language required in order to compete at the NSB is something that cannot be faked by a stage mother. It has to come from within the students themselves.”

I'm sure there are other quotes to be found, but they would just confirm these stories.

The thing about using outliers as the basis for formulating instruction, is that by definition they are "unusual." The conditions by which they got to that point just can't be duplicated for novice learners where the instructional aspiration is much lower than spelling words that few other people can spell.

KDeRosa said...

Yes, much mischief has been perpetrated in the mistaken belief that the lower half of the curve should be taught the same way as the the right tail.

Dick, is that the guy who taught the chimp sign language?

Anonymous said...

Yes. Terrace is the Mentor of Nim the chimp. As I recall, he demonstrated "errorless" learning with pigeons and then moved up the ladder to chimps as Director of the Primate Lab at Columbia.

Libby Maxim said...

i think you do the lower half of the classroom a disservice in thinking that they are not as motivated to learn to spell as the spelling bee champs, raising the bar for the lower half is where we all fall short

spelling is spelling no matter the child, and whatever motivates those spelling bee champs needs to be duplicated for all kids

Anonymous said...

Makes me think of history classes: in schools, we're often told to memorize dates and places either with no context or with generic, boring context. After we graduate and we start to better understand the significance of events and hear some of the more accurate (and colorful) explanations of what really happened, it's a lot easier to remember.

Makes sense that spelling is easier when students are comfortable with word origins, etc.

Unknown said...

I have a theory based on my experience with my expert memorizer/bad speller/struggling reader who now reads well . . .

Some people are really good at remembering things at the meaning level and so they don't remember things below the meaning level, the parts that build up to the meaning level. So my dd10 could memorize whole passages of books but not be able to read the simplest words. She can memorize speeches I give to my little kids--verbatim, and spew it out weeks later.

With spelling, she remembers the words and the meaning of the words and passages--but not the parts (the spellings) that don't have meaning beyond their correspondence to sounds.

I think Dan Willingham's comments about memory in the same issue are apropos . . . we remember what we think about. Some kids have trouble/don't like to think about parts. A champion speller LOVES to think about the parts of words. While it may be difficult to teach a child to love to think about spelling, we can organize the teaching of spelling to give the kids with difficulty a fighting chance at it.

The activities that have worked for my daughter are:

1. Spelling lists based on the sound-spelling correspondences. So a list of words where the sound /ee/ is spelled 'ea' (teach, each, reach, beach, tear (as in from the eyes), dear, beam, seam, cream, etc.)

2. The words are analyzed for their parts, letter teams underlined, and strange spellings circled. She says the sounds for the parts as the words are written multiple times, currently 5 times. If possible, she exaggerates the pronunciation (for example, /lem-On/)

3. She composes original sentences, trying to combine as many words into a sentence as she can (Each teacher sat on the beach eating cream.) She says the sounds as she writes the words.

4. The words are tested until she gets them right at least 3 times in a row. The tests are not lists but dictated sentences I make up using the words from her list and previous words. I test every day. This takes about 20 min. a day (She is 10 and has 15-20 words a week)

5. The words she gets wrong go through the study process each day.

6. Every Friday we do a random, cumulative test from previous lists.

I pick the correspondence to work on based on her writing, then put together a list of similar words.


Melissa
Minnesota

Stephen Downes said...

> Sounds like you agree with Engelmann's theory of instruction. Am I right?

No - but the reason for that is complex. It's not simply that the program is too rigid and not simply that it emphasizes basic skills.

First of all, the method described - whereby a student is shown numerous examples, asked focused questions, and given immediate feedback - is undeniably effective.

We don't need to refer to studies to know this. The ability of children to master complex computer games very quickly is evidence of this. Computer games and direct instruction contain the same basic elements.

This also explains the results of the study. If you ran a set of students through a program of playing Myst, we would find afterward that these students were considerably better at playing Myst than students who were given other tasks.

So the issue - for me - resolves into one where it is not so much the mechanism I am concerned about, but rather the deployment.

The Engelmann approach presupposes (something like) an environment of unambiguous concepts, one where these concepts can be known and selected by teachers beforehand, and one where the learning of these concepts will be sufficient to support future learning.

I don't think any of these three presuppositions is true.

First, the idea of unambiguous meanings and perfect instruction is a chimera. Philosophically, it is widely refuted (se., eg. Quine 'On the Indeterminacy of Translation'). What people - even children - see when they see something varies from instance to instance.

Second, even if we can understand concepts this way, there is an infinite number of concepts, and no (apriori) way to know which of those should be taught.

The Engelmann model supposes that the concepts in any discipline are cumulative - that one builds on the next. (That's how video games work too). However, knowledge is not structured that way. Even in ordinary every discourse a person is faced with a welt of contradictory ideas, meanings, representations and phenomena.

What the student needs to learn is how to construct generalizations ('generalizations' is too narrow a word but I'll use it for now) even in cases where there is not a known or 'correct' interpretation.

Third, a mechanism such as Engelmann's might get children started with some basic tools - such as the ability to project or to interpolate - but it will leave them unable to derive new tools or to deal with new forms of data.

Thus, the responses given to the students must reflect, not a set of prior selections from the instructor, but rather, actual responses from a complex environment.

Such actual responses will display a degree of ambiguity not found in direct instruction. Keeping up the pace and model of interactions, however, allows a student to develop his or her means of responding to that environment.

So - in summary - I think that the Engelmann methods have a lot to recommend them, but that the model has to be adapted to epistemological realities.

Giving people a simple understanding of a complex world will help them pass tests that measure this simple knowledge, but will not give them tools to manage in the complex world.

Anonymous said...

Perhaps data reduction is exactly what is needed to simply a complex world.....

Tracy W said...

Downes: The Engelmann approach presupposes (something like) an environment of unambiguous concepts, one where these concepts can be known and selected by teachers beforehand, and one where the learning of these concepts will be sufficient to support future learning.

I don't think any of these three presuppositions is true.


Okay, this is a bit late, but I just noticed that, and got to thinking about it.

If you think this, then how do you think that good-quality education is possible? This is a serious question. For a start, unambiguous concepts, if you are teaching ambiguous concepts then how do you know if the student has learned them? For effective teaching, surely it is necessary to have an idea of whether the teacher and student have been successful?

If a teacher can't know and select what concepts they are planning to teach beforehand, how do you have effective teaching?

Let's take the example of teaching driving. Now it is generally understood that there are some skills that are necessary to

As for learning those concepts will be sufficient to support future learning, I am not quite sure what this means. Humans are built to learn, as long as we continue to have functioning brains, we will continue to be able to learn, regardless of whether we go to school or not and regardless of how well or badly we are taught at school.

Your disagreement with these three presuppositions makes me wonder - how do you think that, say, reading English could be taught? It strikes me that to teach a kid to read English requires, firstly, knowing how to tell if someone can read English or can't and secondly, knowing what skills/concepts must be taught (eg sound-letter correspondences). Finally, while a kid can go on learning without ever learning to read English (as plenty of illiterate hunter-gatherers can show), one of the reasons for learning to read English is that it supports future learning (eg, from newspapers, from the tax department, from being able to find a course on Pilates, etc). Learning to read English may not be sufficient in the mathematical sense to support future learning, but it certainly is important.

Perhaps there is some ambiguity in the word "unambiguous"? Everything may be ambiguous to a strict philosopher, but that doesn't mean that everything is ambiguous to the same extent. For example, "learning to read English well enough to read The New York Times" is a lot less ambiguous than "Learning to harmonise with the undying circles of nature". And less ambiguous definitions can be refined as necessary down to a point where people can consistently apply them.

I think that it is easy to make too much of theories like Quine's. Yes, no statement of language meets a mathematical level of certainty. But in the real world, practical people like engineers, pilots, sailors, etc, often manage to transmit knowledge to each other faithfully, in areas where evidence of mistranslation is extremely obvious (eg the engine blows up, the plane crashes, the ship doesn't find the Tongan islands). Mistakes still happen, but the rate of mistakes can be reduced by taking measures to reduce ambiguity.

And even if perfect instruction is a chimera, that doesn't mean it's useless. Engineers know they can't reach 100% efficiency, but that's a useful goal to aim towards, and is used incredibly frequently to judge current efficiency.

You go on to say that there is no apriori way to know which of those should be taught. Well there may not be an apriori way, but in a democracy the Government has the right to define what should be taught (as elected representatives of the taxpayers who pay for it). And around the world Governments are quite capable of defining curriculae.

The Engelmann model supposes that the concepts in any discipline are cumulative - that one builds on the next. (That's how video games work too). However, knowledge is not structured that way.

I think you are muddling up how knowledge is structured versus how it should best be taught. The advantage of teaching concepts in any discipline in a cumulative way is that the student naturally gets practice in the earlier concepts as they learn the future ones. This aids memory, without the dullness of practice as pure drill.

What the student needs to learn is how to construct generalizations ('generalizations' is too narrow a word but I'll use it for now) even in cases where there is not a known or 'correct' interpretation.

This strikes me as dangerous. If you don't know what the correct interpretation is, then it's time to be bloody careful about constructing a generalisation. Ideally, you should construct multiple generalisations and test them. If you can't do that, then I'd prefer not to construct a generalisation, and just living with not knowing what is going on.

Third, a mechanism such as Engelmann's might get children started with some basic tools - such as the ability to project or to interpolate - but it will leave them unable to derive new tools or to deal with new forms of data.

Huh? Humans have been deriving new tools and dealing with new forms of data long before formal education. How could a teaching method derive them of this ability? If DI included brain surgery I could believe this, or if it included immediately punishing any kid who dealt with new forms of data or derived new tools, but DI doesn't. Therefore how can you conclude that it leaves kids unable to do such things? There may be a mechanism by which it does do this, but it's certainly not obvious.

Thus, the responses given to the students must reflect, not a set of prior selections from the instructor, but rather, actual responses from a complex environment.

I don't see how this proposition follows from the premise previously stated in your preceeding paragraph. And not just because your premise strikes me as highly unlikely in the first place. Say that children are unable to derive new tools or deal with new forms of data. How would providing actual responses from a complex environment improve their ability to do so? To take an extreme case, engineers and physicists have been unable to break the Laws of Thermodynamics, despite ample empirical work, which of course includes actual responses from a complex environment.

It strikes me as plausible that actual responses from a complex environment could be so difficult as to make it more difficult for people to derive new tools or deal with new forms of data. May I refer you to "The Logic of Failure"? This book supplies information about people's learning from a variety of simulations. One of the general results is that increasing complexity, even by something as simple as introducing a time lag, reduces most people's ability to deal effectively with new data, and instead has a tendency to create superstitions. Now admittedly this is simulations, not real world, but if someone's ability to learn decreases as a simulation gets more complex, why should giving them actual responses from a complex environment be a good teaching method?

This is not evidence that your conclusion is wrong, all I am arguing here is that your conclusion does not follow from your premise.

Such actual responses will display a degree of ambiguity not found in direct instruction.

This depends what is doing the responding. For example, touching a hot stove is most definitely not an ambiguous response for anyone with a fully functional nervous system.

Giving people a simple understanding of a complex world will help them pass tests that measure this simple knowledge, but will not give them tools to manage in the complex world.

This statement is wrong, at least as a universal rule. Newton's Laws of Physics and calculus drastically simplified our understanding of a complex world, and this has given engineers all sorts of tools to manage in this complex world, and these tools actually work in the real, complex world. One of the ways that science procedes is by figuring out simple ways to understand our complex world, by identifying and making obvious the underlying structure.