# Infernal Inference

I’ve always found it much harder to teach how to infer meaning than how to decode. With decoding there’s a clear routemap. With some children, it takes a bit longer to travel that route, but if you stick to the path, in the end you will get there.[1]  But with inference…some children just get it and some really, really don’t.  And when they don’t, it’s really hard to move them on. You show them some text, you point out what you can infer from what you’ve read, they nod (on a good day) but when you ask them to have a go, they are just as clueless. It doesn’t seem to stick. Why is this?

Daniel Willingham, the cognitive scientist explains why. It seems that inference as a skill doesn’t really exist.  No wonder it’s so hard to teach! It’s a bit of a shame then that in the KS reading SATS paper between 16%-50% of the marks are allowed to come from ‘explaining and justifying inferences with evidence from the text.’   In 2016, 36% of the total marks came from questions that involved inference (18 marks out of 50).  I’m not alone in finding inference an area where children struggle.  Nationally, if we just consider the marks available for inference, the average score was just under 9 marks, or 49%.  So this year, alongside vocabulary, across the nation, improving the ability to infer has become the thing we are all working on.

What does Willingham actually mean then, and how might it help us?  In a previous blogpost I quoted Daniel Willingham on this in some detail. The quotes really stayed with me and I’ve been reflecting on them ever since,  and reading more of his writing, so although this post is covering similar ground, I hope it is still useful. Or interesting. Both would be great! So to recap, Willingham explains that inference is more of a trick than a skill. With a trick, once you know it, you won’t get any better by practising it over and over again. Skills on the other hand, improve with practice.

Willingham says that when we teach inference what we are really doing is teaching children to connect ideas, filling in the missing bits the author has left out.  That’s more or less it all there is to it as a technique; connect one idea, one sentence possibly, with another, by filling in the bits the author has left out. Authors always leave bits out; they don’t explain every last detail in the same way that when we speak, we make certain assumptions about the person listening, assuming they already know stuff and so can join the dots. The difficulty with this when it comes to reading rather than talking, is that when we talk to someone, we monitor if they are understanding us. We will look at the person we are speaking to from time to time to check for signs that they understand.   The person listening will give us some useful feedback, maybe nodding or saying ‘mhm’ or ‘uh-huh’; what linguists call giving ‘acceptance signals’ to show they get it.  If we have assumed too much and the listener doesn’t understand, they will send us a signal by saying something like ‘huh?’ or by looking baffled. This is our cue to provide more detail.  The thing about books is that they just don’t care if we ‘get’ what they are saying or not. They don’t monitor our acceptance signals, even though we might nod along as we read a recipe or set of instructions and they certainly don’t rephrase what they are saying if we exclaim ‘huh?’

Trick number one: notice if you understand what you are reading

So then, the first thing we need to teach children about inference is their own crucial role in checking they are understanding what the text is saying as they read.  The book isn’t going to stop and repeat itself or explain in more detail if they don’t understand.  Successful readers expect to understand what they read and know what to do when they spot themselves not understanding. We need to explicitly teach children to check that what they are reading makes sense to them; to nod along as they read in the same way we nod along when someone is talking to us. If something doesn’t make sense they should say ‘huh?’ to themselves and stop and reread the last sentence or two more slowly, to see if that helps. If someone was speaking, we’d ask them to repeat what they had said. Unfortunately, a book can’t rephrase was is written, but if we reread, we might find the bit that is tricky and then be able to make more sense of it.

This ‘expecting a text to make sense’ business seems obvious, but have you ever come across kids for whom being asked to read seems to be like pure torture, who stare grimly at the page when forced, avoiding engaging with the actual words, seeking any and every opportunity to be distracted?  Children who hate reading, even if they appear perfectly able to decode. It may well be that they don’t really get that it is meant to be an enjoyable experience.  I suspect that when children make the transition from reading out loud to an adult to being expected to read ‘for pleasure’ in their heads, a certain percentage just don’t realise that successful reading requires two things. Not only do you have to read the actual words but you also have to check that these strings of words make sense to you. And if they don’t, the reader has a responsibility to stop, go back and reread to try and fathom out what the writer is trying to say. Sometimes this will mean going back two or three sentences, maybe more than once.  Sometimes we might need to reread a whole paragraph again.  If children have never seen this modelled by a teacher, then how would they know that this is what they should also be doing?  Teachers need to use ‘thinking out loud’ to show how they check for meaning as they read. They need to model stumbling over phrasing or meaning and then stopping and rereading, to clarify the sense.  Children need to understand that this is a normal part of being a reader – something expert readers do all the time.

Trick number two: fill in the gaps by connecting ideas

The second thing we need to teach children about inference is that because writers leave bits out, they expect readers to join the dots and connect ideas for themselves. It would be beyond tedious if writers explained everything in minute detail.  We are so used to this as expert readers that we don’t even notice the gaps; that the sentences are connected is obvious to us.   Willingham illustrates this with this trio of sentences.  Bill came to my house yesterday. He dropped a cup of coffee. My rug is a mess.  The first connection the reader needs to make is one of coherence inference; the reader needs to connect ‘he’ with Bill.   Then there is elaborative inference; the reader is expected to draw on their life experience and general knowledge to connect the three sentences together.   In this case, the reader is expected to make the connection that the cup of coffee that Bill dropped is the cause of the mess on the rug even thought this is not stated. It is also assumed that the reader knows that coffee is dark liquid likely to stain a rug and that such stains are not usually desirable. The reader will need to access their long term memory and retrieve information about coffee, rugs, stains and mess; not just what these things are, though that too is important, but how they inter-relate. The writer is assuming the reader can make the following connections: dropping a cup of coffee would cause this dark brown liquid to escape onto the floor. Sometimes floors are covered with rugs. If the coffee fell onto the rug, its colour would remain – stain the rug – even after the rug had been cleaned. People usually clean up after spillages so when the person says that the rug is a mess, they are most likely referring to the stain as being ‘the mess’ rather than the cup and coffee themselves as it is probable that they have already tidied up the mess caused by the cup and liquid. This is because Bill came yesterday. That means the writer is writing about something that happened a day ago. People do not usually leave a mess like that for a whole day, they usually clean it up straight away. This is so obvious to us we don’t even notice we are making these connections and are baffled when children fail to make them.

Teachers need to help children understand how to do this by ‘thinking out loud’ as they read aloud to pupils, by asking themselves questions to show how they are monitoring their own understanding and seeking to make connections between different elements of the text, thus making explicit the thought processes involved in making inferences.

Global inference

The situation can be even more complicated with good stories.  Sometimes the writer will deliberately leave loads of information out.  In fact, it’s the leaving out of key bits of information that makes them so good. For example, let’s look at the beginning of the book The Imaginary by A.F Harrold and add in the extra information the writer assumes the reader already knows.

Being dead is a bad thing. When you are dead you don’t exist anymore and people who like you are usually very sad about that. Most people don’t want to be dead and don’t want their friends or family to be dead either. Most people don’t even like it if strangers are dead. Amanda is a girl’s name so the dead person is female.  I have used the past continuous (was dead)  so that you realise the death has occurred quite recently. I’ve written this as a short, stark sentence to help you appreciate the shock value of this sentence to the other characters, who you haven’t met yet. I’m hoping it will shock you too.  I deliberately haven’t told you if this sentence is the narrator explaining or a character thinking. You will have to read more to find that out.

The words were like a hole in his chest, like a well he was falling down.

By ‘The words’ I mean the previous phrase ‘Amanda was dead’.  The person indicated by the use of the male pronoun ‘he’ is very upset by these words. I have given you two word-pictures to help you understand just how upset he is. When I wrote ‘like a hole is his chest’ I didn’t mean there actually was a real hole in his chest, I was using figurative language to show that it felt really bad emotionally, in the same way that a hole in the chest would feel really bad physically. In a similar way, there is no actual well and ‘he’ is not actually falling. Again, this is a word picture to try and describe the emotional turmoil the male is undergoing as a result of hearing about the female Amanda’s death. Although strangers usually feel a bit sad when hearing about someone’s death, the figurative language shows that the sadness is very strong. Therefore, there must have been a strong emotional attachment of some sort between Amanda and ‘he’. Finding out about this attachment and why she is dead is going to be explained to you in the rest of this story. You can also expect to find out how ‘he’ gets over his strong feelings. This is a children’s story. Probably it will end happily. Don’t worry that you do not yet know who ‘he’ is; I will explain that to you at some point. Not telling you is a way of making you want to read on, to find out that information.

Writers – thank God – leave stuff out and expect us to fill in the gaps. Children need to understand this. So when they don’t understand, they know they need ask themselves what is it that the writer is assuming that I know that I don’t.  Or maybe the writer is just playing with me…

The joy of inference

If children are going to learn to love stories, they need to understand that the author sometimes deliberately leaves out information and it is the trying to work out what has been left out that creates the frisson, the excitement that makes reading stories pleasurable.  Why is Amanda dead? Who is this male protagonist? What will he do next? Will he ever recover? From the first two lines, the deliberate ambiguity makes us want to resolve our ignorance and read on.  So readers need to appreciate that sometimes the author is teasing us, dropping hints and clues but never quite enough information because it’s the not-knowing that makes reading fun.  That’s why some books are described as real page turners.  Readers have different tolerances for this not-knowing. A whodunit is all about the journey from not knowing to knowing, some poetry can remain elusively ambiguous even after several readings.  Some readers like this; others really don’t. The satisfying closure of an unambiguous – if unexpected ending – is what for many readers makes a great read.

Global inference is the kind of inference needed to make connections and predictions across whole texts. For this to happen, children need to learn about text conventions. For example, to return to The Imaginary, the fact that the author does not tell us immediately about why Amanda is dead or what her connection to ‘he’ is  is  confusing, but it is a deliberately planned confusion, to make us want to read more to resolve the tension of not knowing.  As I’ve mentioned in a previous blog, the six signposts in Notice and Note are great ways of teaching children to look out for subtle hints from the author about the way a story is going to go. When Willingham talks about inference, he is not really talking about this kind of global, whole text inference.  This involves learning about how authors craft character, use conflict to drive plot and flesh out themes and use foreshadowing to hint at later developments. These do need to be taught and deserve quality curriculum time devoted to them. These are not what his research was looking at. Indeed, his point is, if we spend too much time teaching about basic inference, we won’t have time to do the important stuff. Such as…

Developing Vocabulary

It is hard to make connections if you don’t know what some of the words mean, so teachers need to develop children’s vocabularies.    I’ve written before about ways of doing this so I won’t repeat that here, beyond mentioning that if you haven’t yet read Isabel Beck’s Bringing Words to Life then you are missing a trick. To state the obvious, children need to encounter lots of unfamiliar words in meaningful contexts, especially the kinds of words that occur more frequently in books than in the spoken word.  Children need to be read to a lot and to read to themselves a lot and to enjoy reading so much that they want to do more and more of it.

The inference problem is really a knowledge problem

Willingham wants teachers to know about the research on inference.   This says that while teaching children how to infer is quite important, there is little benefit to be had in teaching this for more than a few lessons because the techniques of monitoring ones understanding and then trying to make connections are easily learnt within a few hours of instruction. No additional benefit is gained by spending any more time on mastering these techniques beyond this.  To which most teachers I know would respond ‘easily learnt my ****!’   as inference does not seem to come easily.  But the meta-analysis is clear; once you know to monitor your understanding and to make connections and have spent a few hours working on that, further practice does not make you any better and is time wasted.  That’s because once you know these tricks, what will get in the way of understanding texts is gaps in your life experience, general knowledge and vocabulary. You will realise you don’t understand something and try and make the necessary connections, but still fail because you don’t have the knowledge to know how the things you are trying to connect actually relate.  Now the very same teachers nod in agreement; it’s not that children don’t know they should try and make connections; they just don’t know what the connection is on this occasion. Or the next. Or the one after that.

I’ve used this example before; it’s from an old reading sats paper.

The old bear roared in frustration and waved at the empty air with his huge paws, then reared up on his hind legs .’

The children I was teaching didn’t know the meaning of the words ‘frustration; ‘paw, ‘reared up’ or ‘hind’.    They found the term ‘empty air’ confusing.  They were meant to be inferring why the bear ‘reared up.’  We didn’t get very far.  It was not the most successful lesson I’ve ever taught.  All the teaching in the world about monitoring your understanding and trying to connect ideas won’t help when so many words are unknown.  So instead of spending too much time on teaching children how to infer – Willingham says about 10 hours is enough – acquiring a broad vocabulary and a rich base of background knowledge will yield more substantial and longer-term benefits.’  As children need to be fluent decoders before teaching the tricks of inference is really pertinent, Willingham concedes that since children develop at different rates, a little bit of inference and other reading comprehension teaching in the fourth, fifth and sixth grade (years 5, 6 and 7 in the UK system) is a sensible use of time, but beyond that, it is time wasted.

It is interesting that the recent Education Endowment Foundation Guidance Report on improving literacy at key stage 2 hardly mentions developing this ‘rich base of background’ knowledge at all, beyond including Scarborough’s Reading Rope, where background knowledge is the first strand mentioned (not because it is top priority, I think the layout of the strands within the ‘language comprehension aspect is fairly arbitrary and it is just chance that background knowledge was placed on top).

The report says that the Reading Rope model can be used as a diagnostic tool to identify areas to focus on. What Willingham is cautioning us against is over-diagnosing problems with inference at the expense of the real culprits; lack of background knowledge and under-developed vocabulary.  In the report’s section on teaching comprehension strategies, it suggests it is useful to ‘activate prior knowledge.’  But the crucial point missed here is that for many of our pupils who struggle to understand texts, they don’t have this ‘prior knowledge’ to activate in the first place! In fact, the same section recommends ‘extensive practice’ of exactly the sort of specific strategies that Willingham warns should only be used for short periods of time, as further practice once the strategy is learnt is fruitless.

This is very scary for primary headteachers and ks2 teachers who are held accountable via a reading comprehension text, for it is saying that what we are mainly being tested on is our students’ vocabulary and background knowledge.   But since we never know in advance which knowledge and vocabulary our pupils will be tested on – white giraffes, warthogs and dodos last year, who knows what in a week’s time[2] – how can we make sure they know enough to do well?

This is where we need to take deep breaths and be brave. We are here to teach children to be good readers for the rest of their lives, not to teach children to appear to be good readers one Monday in May.  Sure, we’d be foolish not to spend a fair bit of time in the immediate run up to the tests practising test technique.[3] But we can’t let the stupidity that is the reading comprehension test warp the whole of the way we teach reading in ks2.

If having rich background knowledge is key to becoming a successful reader, then accumulating that knowledge needs to start as early as possible.  We need more non-fiction reading aloud to children in the early years and key stage1, alongside fabulous stories. The emphasis on spoken language in the early years is crucial, but we must remember to give them interesting, memorable stuff to talk about; not just great stories.  The ‘understanding the world’ strand should be the driving force underpinning the early years’ curriculum.  Practitioners should know what vocabulary and general knowledge children will have acquired as a result of their topic on ‘pirates’, for example.  Particularly in a context where children are disadvantaged and may have limited life experienced, we need to expand children’s horizons in all year groups, taking them on trips, bringing visitors into school and fully appreciating just how urgently our children need to acquire the kind of general knowledge middle class children are routinely exposed to at home. We need to understand that the humanities and science lessons are the main places where children will gain this knowledge and make sure our curriculum and pedagogy ensure that knowledge is imparted and retained.  Those children should have known what a paw was and what hind leg means. Cutting back on foundation subjects to improve reading is a false economy.

We should  also devote prime curriculum time to reading amazing books to children, expanding their vocabularies and doing all we can to get them to want to read for themselves. We need great libraries, teachers who are knowledgeable about children’s literature and who can help every child find the book that they will fall in love with; time in the curriculum for both reading to children and for children to read to themselves. And we need lots and lots of books.

[1] In 99.9% of the time. I’ll concede that for maybe 0.1% of children you eventually need to try another approach. Stressing eventually here.

[2] I’m writing this the week before the 2017 key stage 2 reading sats test. Indeed, this time in a week, the completed tests will be safely in their grey plastic bag, awaiting collection. *gulp*

[3] I saw a hilarious lesson last week where the teacher told them that the reason instructions to ‘tick ONE box’  were written in bold was because the markers were fed up with children sometimes ticking two, so the bold was the marker shouting at us to make us remember.