Engrade also integrates enough built-ins (messaging, discussions, quizzes) that it could be used as an LMS alternative–especially for those who dislike the idea of a totalitarian LMS but need basic class management tools. (If you’ve used Engrade as a LMS, let me know how it went!)

And while Engrade will probably suit the needs of higher ed instructors, features such as “Attendance”, “Seating”, and “Discipline” seem clearly targeted at K-12 scenarios.

• Carver identifies 5 “gears” of reading: scanning, skimming, rauding, learning, memorizing
• Lunzer identifies four processes: scanning, skimming, receptive, reflective
• O’Hara recognizes Lunzer’s four processes, and identifies 12 types of reading activities that overlap with Carver’s 5 gears
• Birketts recognizes vertical (deep) and horizontal (range) axis; Levy calls these intensive and extensive, and predicts hyperextensive
• Wolf refers to any reading apparently above skimming as deep reading
• Guthrie suggests that scanning is not a type of reading, arguing that information finding is a distinct cognitive process that utilizes reading, but deserves its own model

I’m still synthesizing the meaning of these distinctions and relationships of the categories to cognition. For instance, depth suggests completeness of coverage of the text, and reflectiveness suggests degrees of attentiveness, even interaction with the text. Both of these are process variables. So the gear of learning, for example, which aims to be deep and attentive, suggests increased demands on cognitive resources such as working memory. Speed is a reader variable, but it is also limited by the reading strategy or type as an aspect of process, e.g. skimming is meant to be faster that rauding, while memorizing is at least initially slower. Intensive reading, such as learning, is presumably slower than extensive reading for pleasure. I haven’t yet looked at how these process variables affect outcomes, though I have seen such studies referenced in the literature.

I’ve attempted to visually approximate the relationship of Carver’s gears of reading to the following aspects of process: depth / completeness, attention, and speed:

The “measurements” of these areas are speculative, but might eventually be teased out by research. I do not have much experience visualizing relationships between variables beyond basic, two-dimensional charts, so if you have other ideas, please share.

The real purpose of this exercise is to begin thinking about types of reading and their relationship to cognitive processes and models.

A quick search on Google shows that David Walsh has already worked out a very elegant script that approximates this. My own idea would be to not use a “flashlight” effect, but a “frosted window” effect using moveable, translucent PNG layers (rather than a background) that retains basic shapes, features, and colors, but obscures microfeatures–i.e. the graphemes–of text. Here’s one mock-up to illustrate the idea:

Now there are a number of obvious problems with this idea in research practice–the obfuscation may significantly alter the layout and (potentially distracting) elements of the page; it may negatively affect users ability to gauge document length and position; it wouldn’t necessary reflect normal reading behavior; and, of course, it could be terribly annoying to the reader, etc. That’s just off the top of my head; if you have more ideas, please share.

Dual-coding theory suggests that learning can be actually be enhanced when cognitive load is increased in a specific way, such that both auditory and visual coding is activated simultaneous, and in concert. This is only true, dual-coding warns, when the information does not compete. This is where the distinction between verbal and non-verbal becomes important. Two forms of verbal information–visual verbal and auditory verbal–is likely to compete, even if one is spoken and one is heard. Another example of competing information would be visual non-verbal and visual verbal information, but not visual non-verbal and auditory verbal. Though I haven’t seen any studies of it, I daresay because we can easily discern lyrics from a song, non-verbal auditory and verbal auditory do not compete.

As I tried to understand why certain combinations of information compete and others don’t–and indeed are complementary–I turned to a book I’d read earlier this year, Stanislas Dehane’s “Reading in the Brain”. Dehane begins by explains that reading takes advantage of massive parallel processing (I love the daemons/pandemonium metaphor). He then explains the dual-route theory, which postulates that reading proceeds predominantly along one of two routes: phonological (or graphophonological) (visual–>auditory–>lexical) or orthographic (or lexicosemantic) (visual –> lexical), depending on the word being read:

All writing systems oscillate between an accurate representation of sound and the fast transmission of meaning. … Two information processing pathways coexist and supplement each other while we read. When words are very regular, rare, or novel we preferentially process them using a “phonological route”, in which we decipher the letter string, then convert it into pronunciation, and finally attempt to access the meaning of the sound patter (if any). Conversely, when we are confronted with words that are frequent or whose pronunciation is exceptional, our reading takes a direct route that first recovers the identity and meaning of the word and then uses the lexical information to recover it’s pronunciation.
(Dehane 38-39)

Research beginning and the late 80s and proceeding to today have supported this theory. Historic brain imaging experiments in 1989 from Petersen et al showed that different regions of the brain were used when listening to words (superior temporal region) vs. silently reading words (left occipito-temporal region, aka the brain’s letterbox–different from areas of brain which code broader visual stimuli). Some silent reading tasks then transmit to the superior temporal region for coding. (67-74)

Initial input pathway differs depending on whether we hear or read a word, but cortical activitiy later converges onto the same set of language areas.” And appears to extend out (and back, and out again) from there, apparently activating more areas of the brain than seems “necessary” to understand, well beyond the actual recognition of the word. (103-104)

That doesn’t mean we only use auditory areas when listening to words; we also may use them as the phonological route suggested by dual-route theory. This is described by later fMRI-based brain imaging experiments that found…

“Common words, as well as words whose spelling is irregular, head straight for the lexical areas of the middle temporal lobe… Other words, whether they are regular, rare, or unknown, are first pronounced mentally using auditory areas of the superior temporal lobe before a meaning is attached to them.” (116)

In terms of speed, it looks like the direct lexical route is faster (words we know; ~200 milliseconds) that the auditory first route (words we have to sound out; ~400 milliseconds).

Because either may be used, and because the route is largely based on the reader’s prior knowledge, it seems important not to interfere–not even risk interference with whichever route is taken. This supports dual-coding’s rule against competing information (verbal and verbal). It doesn’t complete the picture, however, or, if it does I haven’t quite discerned the layout. One (of many) questions I have: what is the impact on activation time of these two routes if the task is burdened with auditory stimuli, such as narration (either conflicting or coinciding)? I would guess that in instances where the the word is first “pronounced mentally” would be more subject to interference or delay from auditory stimuli or narration.

In order to put cognitive load theory into practice I think a clear distinction between both verbal and non-verbal as well as visual and non-visual is important. The following diagram is my first attempt to clarify these fairly simple distinctions:

Aleven, V, Koedinger, K. (2002). An effective metacognitive strategy: learning by doing and explaining with a computer-based Cognitive Tutor. Cognitive Science 26, 147-179.

In this report the researchers explain how explanation was shown to increase learner performance and increased time on task. In this particular case the learners input their explanations into a computer interface, but I wonder about self-explanation–that is, explanations crafted only as a part of one’s reflection, only for one’s own consumption. I make this distinction because though learners were “communicating” with a non-human (software) I suspect it was clear to the learners that their explanations would be archived and relayed to the instructor. So to me there may be an issue of learner privacy–a privacy that may shield them and allow them to more fully express themselves, or, conversely, a privacy that masks them and allows them to neglect a fuller explanation as might be made under instructor scrutiny.

There are more questions to raise (some fascinating ones regarding instructional design for solitary learning) and I hope to consider and investigate them in the coming weeks.

It’s little more than a stub right now, but I plan to come back to it, and I trust others–especially those in Jon Mott‘s IP&T 692R course, “The 2 Sigma Problem”–will contribute to and expand the article.