100% Coverage ≠ 100% Comprehension

A question by a member of the Latin Best Practices FB group prompted me to look into text coverage, which ultimately led me to comprehension. These are two ideas that a lot of people have misinterpreted, much like the “4%er” figure, and even “90% target language use.” I’m thinking people have a hard time with mathematical concepts, and maybe we should avoid percentages moving forward. But first, we should take care of what damage has already been done by looking at simple examples right away:

Text Coverage
Text coverage is measured by tokens. There are five tokens in the sentence “the bird sees the cat.” Two of the tokens in that sentence happen to be the same word. Therefore, “the” represents 40% text coverage. If the reader doesn’t know “the,” they have a text coverage of 60%. The reader who knows everything except “cat” would have a text coverage of 80%.

Comprehension
Comprehension is a different idea entirely. If the reader who doesn’t know “cat” were asked “what does the bird see?” and it were scored, they’d have a comprehension score of zero. If they were asked two questions about the bird, and two questions about the cat, their score would be 50% comprehension with their 80% coverage of the text. Not the same thing.

Reading
Laufer et al.’s research shows that learners need a text coverage—not comprehension—of 98% ideally to read with ease (and 99-100% whenever possible), but that’s just getting through the reading. That 98% figure is just the start of comprehension.

Hold up.

Yeah, that’s right. Knowing 98% of a text—STOP!!—Remember the first section on tokens. It’s not 98 out of 100 different words, but 98 of 100 tokens (i.e. some words probably repeat). So, knowing 98% of a text doesn’t even guarantee comprehension of what is read. That’s quite the trip, isn’t it? It gets worse when we look at some findings from one of Eric Herman’s Acquisition Classroom Memos on exactly how [in]comprehensible reading can get with what seems like decent text coverage.

There’s a lot in that chart, but compare the text coverage to comprehension scores. Even 95% text coverage can get woefully low comprehension (55%). Keep it mind that the higher scores are still in the “most” range, as in learners are understanding most of what they read when they know 95%+ of a text. Also, those vocabulary sizes are incredibly high for what the majority of K-12 teachers should expect from their students. Eric also adds some context to the research:

“For the most part, the above reading studies were done with high proficiency students, ungraded and academic texts, and count word families. A reasonable prediction is that even higher text coverage and vocabulary size numbers are required to enable adequate comprehension of graded texts by lower level proficiency students. And this is not considering levels necessary for a confident and pleasurable reading experience, which would undoubtedly be even higher!

Higher would be 100%. Let’s make sure we set the record straight:

  • Students need to know 98% of a text to read it with ease.
  • Reading with ease from knowing 98% of a text can still result in much lower comprehension scores, like 70%.
  • Coverage ≠ comprehension

Providing students with texts of 98%…even 100% coverage of known words is step zero. It’s actually the minimum hope we could have for students reading with ease with high levels of comprehension. It turns out that text coverage isn’t very important to look at, because even knowing 100% of the words doesn’t guarantee 100% comprehension. It all goes back to vocab as top priority, sheltering whenever possible so gradual exposure to new words increases vocabulary without the burden of incomprehension. What does this mean for class? Probably using even fewer words than you think! Students can’t magically learn thousands of words, so if we expect them to comprehend high levels of what they read—especially during any kind of independent reading—we must use and create texts with a very limited number of words.

90% Target Language Use? How About Your Message Count…

Forget that 90% figure (i.e. 90%+ of all the language provided by the teacher as input-provider should be in the target language)…How many messages are you providing? I did a quick search for Latin lessons:

Here’s the first Google hit for “Latin lesson” with 2 messages, the first recurring 3 different times. The second Google hit contains 0 messages. The third Google hit contains 1 message.

Continue reading

“Getting Students to Speak” & Min/Max Partner Retells

How do we get students to speak the target language?

Provide input.

At least, that’s what no one disputes, though not every teacher does enough of it. The biggest misconception regarding how to get students speaking is based on the assumption that the goal—speaking the target language—must be part of the process. This makes sense, but we don’t have much evidence to suggest this is true, despite how intuitive it seems. In fact, if you want get all Second Language Acquisition (SLA) technical, in 1995 Merrill Swain—herself—called her own Output (i.e. speaking/writing) Hypothesis “somewhat speculative” (p. 125).

Continue reading

NTPRS 2017 Takeaways

Before having the opportunity to present a couple workshops, my mind was blown quite sufficiently during the week. Overall, the Advanced Track with Alina Filipescu and Jason Fritze got me thinking about aaaaaaaall the things I’ve forgotten to do, or stopped doing (for no good reason) over the years. Thankfully, most of them are going to be soooooo easy to [re]implement. As for the others, I’ll pick 2 at a time to add—not replace—until they become automatic. This will probably take the entire year; there’s no rush!

Jason referred to high-leverage strategies—those yielding amazing results with minimal effort (i.e. juice vs. squeeze), and I’m grateful that he called our attention to everything Alina was doing while teaching us Romanian. ce excelent! I’ll indicate some high-leverage strategies, and will go as far as to classify them as “non-negotiable” for my own teaching, using the letters “NN.” I’ll also indicate strategies to update or re-implement with the word “Update!” and those I’d like to try for the first time with the word “New!” I encourage you to give them all a try. Here are the takeaways organized by presenter:

Continue reading

Assumptions & Definitions: Establishing Meaning

1) Our goal is reading Latin via acquisition (universal to humans).
2) Comprehensible Input (CI) is necessary for acquisition.
3) Teaching with CI means providing understandable messages in [Latin].
4) Receiving understandable messages as a student means listening and reading—not being forced to speak [Latin]. N.B. Single word/phrase responses are not considered “forced speech.”

This was the opening slide to my recent CANE presentation on how to continue Teaching with CI after discovering it, largely influenced by the 13-post series on a CI Program Checklist containing many resources and support materials for people just getting into this way of teaching. #3 above deserves more attention. How do we actually make something understandable?

Continue reading