What follows is going to be a written up version of my previous researchED talk.
First disclaimer. This talk operates under the assumption that the pupils in your class have mastered word recognition, which arguably is a skill.
What is a skill?
When we look at the etymology (the history of the word) of the word skill it comes from an Old Norse word meaning knowledge. When we look at an internet definition we get the following…
This is an inoffensive definition of the word, but gets to the crux as to why reading comprehension is not a skill. It all comes down to the word ‘practise’ as it implies an element of transferability. Imagine the footballer who can dribble a ball with great precision on their home astroturf, but fails completely when dribbling on an away pitch made of grass. Only being able to dribble a ball in such narrow confinements would not be of much use. Thankfully there is an element of transfer. Those that can dribble a football with accuracy and precision are not confounded to replicating that in the conditions of which they have learnt it. However, I am arguing that reading comprehension does not work this way. There is very little which we should infer from the comprehension of one age-related text that should then be used as a proxy for comprehension of further age age-related text. Yet, the teaching of reading comprehension (certainly in England and Wales) treats it as a skills-based subject that has transferability. Why is this the case?
I think this is because we (teachers and the general public) are bombard with messaging that it is so. Walk into Waterstones or WHSmith and you will find books like this:
Popular teaching resource sites will make reference to reading ‘skills’. This is not to say that these resources are implying that reading comprehension is a ‘skill’ like dribbling a football, but that they may contribute, implicitly, in doing so.
And Ofsted, in their reports based on the new framework, make reference to comprehension skills across schools with different gradings.
I suspect, however, that if we got all those inspectors together and asked them individually what they meant by comprehension skills that we would get as many answers as the number of inspectors.
And finally we have the worst culprit of all from the STA – the KS1 and 2 content domains.
These outline what the standardised assessments at the end of KS2 (used to be KS1, but those are no more now) attempt to assess through various question types. Not only that, but as you can see with the table on the right it even gives us an approximation of the number and percentage of marks that are available in the KS2 reading assessment. So from Year 3 onwards, the reading diet of pupils could potentially be extracts of text and a focus on getting better at these particular question types.
So naturally, when accountability measures are what they are, bets are made as to how get pupils over the line. Retrieval of information from a text and inferring meaning could make up to 100% of the assessment, so it makes sense that those responsible for the teaching of reading at any level would gravitate to the idea that if we teach these as isolated skills and pupils practise, then they will get better at retrieving and inferring, and this will help them cross that magical 100 scale score boundary. Given that comparison and prediction could only make as much as 6% of the assessment, schools will teach these, but questions of this type they will not come up much, so schools won’t focus too much on these. An interesting experiment to do in your own time is to go to resource sites that provide comprehension questions for you and check the percentages of the question types they ask and compare it to the table on the left above. I imagine that they would be quite similar for obvious reasons.
Out of these domains appeared a whole cottage industry of the reading domain question level analysis (QLA).
Year 6 teachers will be familiar with this. They will do some practise SATs with their class, dutifully complete the QLA (which took ages) and then the excel sheet would tell them which of the domains pupils did not score well at. This data was used as a proxy for who is ‘on track’ to pass their SATs, but was also provided intervention groups (Pupil C and D for retrieving key info). Then we would continue to teach that skill as if all pupils were missing was opportunities for more practise so that they could just get better at it. I remember feeling elated when I found this website early on in my career as it would provide extracts of text with question types based on the skills.
In my own experience of the classroom would happily set an activity like the one of the left thinking that this was going to improve the final outcome on the KS2 reading paper as they were practising that skill of inference.
As alluded to above, I was just as guilty of this when teaching. I do not blame any stakeholder whatsoever for this type of reading pedagogy at all – you don’t know what you don’t know. But we know a lot more now.
Shanahan to the rescue!
At some point I stumbled across Tim Shanahan’s blog. There he perfectly describes this skill based pedagogy, and it certainly struck a chord with the similarities between what was happening in the classroom and his description of what ‘comprehension skills’ are.
Then I read the next part…
And when I thought it couldn’t get any worse…
When I first read this blog, it certainly was a kick in the teeth. Through the way reading comprehension material is marketed to teachers and the general public by businesses, to the wording used by Ofsted in reports and the way that the whole assessment of reading at KS2 had been devised, a pedagogy of teaching reading appeared that lacks any kind of psychological reality.
What is comprehension?
Before we look at what can be done to support pupils with comprehension (assuming all is well with regards to word recognition), it is worth spending a bit of time looking at what we mean by ‘comprehension.’ There are many about, but this one that I always find myself going back to.
Reading comprehension is a “process of simultaneously extracting and constructing meaning through interaction and involvement with written language,” (Snow, 2002).
I this as it emphasises the dual role that the reader has in the process – extracting and constructing meaning. Extracting meaning can be interpreted as word recognition. Constructing meaning implies that another process is happening. One model for that process is the construction-integration (CI) model. A model that I think leaders of reading should know. Let’s focus on ‘construction phase’ first.
Let’s say you see one word in isolation. That word is bark. It exists in the environment and you decode the word successfully. In the CI model, word meaning is not held in long-term memory, it is constructed in working memory using context and prior experiences of words that do exist in your long-term memory. These are demonstrated by the words in blue, but certainly not limited to these words. Other related semantic connection could be activated. The construction phase uses all information about word meanings and syntax that is available in long-term memory. The issue here is that without more context you have no idea which meaning the author is referring to. Meaning is constructed in context. This is where integration comes in.
So let’s add the word ‘dog’. Now more meaning can be extracted as we can integrate context of the sentences with our experiences. When we see ‘dog’ the activation of the semantic words relevant to that possible meaning increase (as indicated by the words in green), and the semantic connections that it seems unlikely to be related to are deactivated. The integration phase selects on those aspects that are contextually relevant.
So far we have seen this for individual words, but it works in a similar vein for sentences with some subtle differences.
Suppose we have the sentence ‘Time flies when you are having fun’. Each word is read and an experience of that word is activated in real time as each word is read. In this case, experiences of clocks, actually flies and other experiences of time and flight are activated.
Reading the rest of the sentence activates other previous experiences.
At this point, we break down the sentence into what Kitsch describes as its propositional structure. This structure is just means of extracting, usually through rewriting, a single literal meaning from text. From the sentence above, many literal (not necessarily accurate or correct) meanings can be found. ‘Time flies’ could be interpreted as type of fly that may only appear when you have fun; as a clock literally flying away but only when you have fun or as a metaphor to describe the passing of time when one experiences enjoyment. Right now we have linked the text base with our prior knowledge. The next thing we do is stabilise this network through the integration phrase.
As above, we now use the context to extract the meaning of this sentence. More text may be available to do this, or some additional information in the environment may provide what is needed. However it happens, it is crucial that the correct meaning that the author intended to is reached. Obviously this happens an incredible speeds when actually reading, but I hope this break down helps to understand what may be happening.
At the paragraph level much of the same happens, readers just need to be aware of how previous sentences (and the propositions they have created) link to what is currently being read.
This model is useful as it has some explanatory power as the process of comprehending:
- Experiences of words are activated as they are read individually.
- Those experiences are kept in long-term memory, and each pupil will have different experiences
- We not remember word for word what we read. We break a text down into proportions.
- Meaning is created in working-memory using context from the text and overlaying it with our prior experiences in long-term memory.
- Successful comprehension is the of extracting and construction of meaning as the author intended it to be.
This can then provide a theoretical framework why a ‘read all of it and answer the questions’ approach to comprehension may not work for pupils. If meaning is created after each word and we update our CI model accordingly after each word, then pupils may believe that they have complete understanding of text when in fact they do not. As a teacher, when pupils do ‘read it and answer the questions’ you do not know at what point in the text a pupils CI model has created an incorrect meaning.
This makes intervening that much more difficult. If a pupil’s understanding of plate boundaries means the literal boundary between plates that you put food on, then understanding what a fault line is in relation to actual plate boundaries is unlikely to happen. This will then have implications for comprehending the next paragraph. My preference is to mostly stop this type of exercise, but this may be beyond the control of teachers at an individual level. My suggestion then is rather than let pupils just read a text and answer the question, split the text and questions into their relevant sections and adopt a ‘read, answer, check and repeat’ mode. The pupils read the related paragraph, they answer one question and the answer is checked. This at least makes sure that ALL pupils are building up an accurate understanding of the text.
A quick note on background knowledge
Clearly, background knowledge is important. It is the cornerstone of the CI mode. A recent review of background knowledge mentions that it has a range of effects on reading. This includes those with high background knowledge of a text being able to follow along with texts that were less cohesive, to having better quality ideas of what is happening in the text. But we should not forget that background knowledge does not have to equate to content knowledge of the text. There are other aspects of knowledge that can be taught (and could be considered background knowledge in a certain sense), and these will be explored in part two.
Mr Almond: Nuts About Teaching 