Showing posts sorted by relevance for query inference. Sort by date Show all posts
Showing posts sorted by relevance for query inference. Sort by date Show all posts

Saturday, 27 January 2018

Reading Comprehension: A Structured Way Of Teaching Inference-Making


Following on from my blog post entitled 'Research Findings: Are There Different Skills Within Inference?' this blog post provides practical advice about how inference-making might be taught in a structured and simple way. If you are interested in the research base for what I put forward in this blog post then do read 'Research Findings: Are There Different Skills Within Inference?' first. 

Based on my reading of research, and on my analysis of my own experience of teaching inference-making, I put forward that when we make inferences we are thinking about why things are or are not so. To break that down I also suggest that:

  • Inference-making can revolve around actions such as what is (or is not) said, done, thought, felt, believed, perceived and so on. Inference-making is also about why, how, when and where these actions take place. These action-based inferences might pertain to actions in the past or the present, or to intended actions.
  • Inference-making might also involve events (or happenings, or occurrences) such as what happens or does not happen. Inference-making is also about why, how, when and where these occurrences take place. There are obvious crossovers here with how the actions of a story's characters, or a text's subject, influence events. Making inferences about events might focus more on things that happen with or without human influence e.g. as a result of natural processes or a sequence of other events.
  • Thirdly, inference-making might be about the state of things. These inferences might refer more to inert, insentient things such as places, buildings and objects and could focus on why things are as they are, what they are, how they came to be and where they are. In a way, Class 8: Instantiation of Noun (see my previous blog post) is actually an example of this. Using the same example, and posing is as a question that a reader might ask themselves: What is breakfast? It is bacon and eggs. In the same way Class 1: Referential is an example of making an inference about a thing's state: What is it? It is a fork.

If at least the majority of inferences that we make whilst reading revolve around a verb, including forms of the verb 'to be', then we have a sensible starting point to teaching inference-making. There may be disagreement about whether or not inference can be learned, but what is certain is that it can be modelled by teachers and practised by children in simple comprehension lessons where questions are posed and answered. Since much of reading instruction follows this process it would make sense to be a little more deliberate about teaching inference-making, especially as it is not always easy to do - the very nature of it means that information is not always explicit and takes more finding.

Reading lessons involving comprehension questions, I'd be willing to bet, often follow one of the two patterns:
  • a sequence of questions that naturally arise from the text, usually a mixture of different reading skills
  • a more deliberate set of questions that aim to allow children to practise a specific set of reading skills
These are fine if your aim is for children to have a complete understanding of a piece of text, or if you are giving children the chance to practise a range of skills after they've had specific skills teaching. What I suggest, at least for lessons where you intend to teach reading skills, is that individual skills are modelled by the teacher and practised by the children. If, for example, you wanted children to get better at making inferences you would model inference-making and then make provision for children to practise inference-making.

But, even this presents a problem: not all inferences are the same. A teacher might model an inference about why something happened and then give children practise questions about how someone feels. This won't allow a child to practise particular skills; only children who are already very skilled in making inferences will be able to answer them and in this case the child would need some more challenging work.

Children will probably benefit best from having the chance to practise specific inferences based on the different types of inference-making listed above. If teachers can provide questions that are of a similar structure, and provide structures for the answers too, then children who are at the learning stage of inference-making might have a better of chance of being able to make inferences whenever they are reading.


This blog post is the second in a series of three:

Part 1: Research Findings: Are There Different Skills Within Inference?
Part 3: Questions To Ask When Teaching Inference-Making

See also:

Scaffolding Inference - my blog post about how first asking vocabulary and retrieval questions can guide children towards answering inference questions.

Monday, 5 February 2018

How To Use Questioning When Teaching Inference-Making


How To Use Questioning When Teaching Inference-Making
In my last blog post on inference-making (Questions To Ask When Teaching Inference-Making) I provided lots of questions which might support inference-making, along with some suggested answer structures for teachers and children to use when answering inference questions. In this blog post we will look at how these questions can be used wisely in lessons so that children's inference-making skills are developed.

Anne Kispal, in her literature review entitled 'Effective Teaching of Inference Skills for Reading', writes: "Underpinning the research reviewed is the assumption that pupils must be explicitly taught the skills they need for comprehension. They cannot be left to pick them up through simple exposure to texts, or through the natural process of maturation." (page 24) It is clear that we should teach children the strategies they need in order to be able to understand what they read - the strategy we are concerned with explicitly teaching here is inference-making.

The questions I shared previously should be used carefully - they are not solely for use in a written comprehension activity which children complete independently. They should also be modelled, discussed, answered orally and asked about aurally-presented texts as well as read texts.

Before we get into the nitty-gritty of the why, I propose a sequence (flexible, of course) to help use inference questions in the most effective way:
  1. Teacher reads aloud a part of the text
  2. Teacher allows children to read the same part of the text
  3. Teacher provides a summary of the text
  4. Teacher models inference-making (which might include clarifying word meanings, locating specific information and discussing necessary prior/background knowledge)
  5. Teacher provides a second summary of the text which takes what has been inferred into consideration
  6. Teacher reads aloud next part of the text
  7. Children read next part of the text themselves
  8. Children summarise text
  9. Children answer inference questions (and any supporting vocabulary, retrieval and background knowledge questions, this could be a written task, or an oral one)
  10. Children summarise text a second time taking into consideration what has been inferred
  11. Teacher models answers and, if written, children edit their work to improve their answers
Now let's see a break down of why it might be a good idea to roughly follow this sequence when using the inference questions:

1. Teacher reads aloud a part of the text

On the Reading Rockets website (a great and accessible online resource) Judith Gold and Akimi Gibson provide an excellent summary of the research on reading aloud:

"Reading aloud is the foundation for literacy development. It is the single most important activity for reading success (Bredekamp, Copple, & Neuman, 2000). It provides children with a demonstration of phrased, fluent reading (Fountas & Pinnell, 1996). It reveals the rewards of reading, and develops the listener's interest in books and desire to be a reader (Mooney, 1990)."

Whether or not you have the children reading along with you is another matter; David Didau, in his blog post 'The Problem With 'Reading Along'', proposes that we don't because the act of listening and reading at the same time can overload the working memory and hinder comprehension of the text. If that is true, then the next step is an important replacement for children reading along.

Note: during this read-through it is best not to stop reading to ask too many questions. Although Kispal summarises that teachers should "practise inferential questions on aurally presented texts" she also provides these cautionary notes on questioning:
  • not to interrupt pupils by asking questions during reading time
  • not to launch into questioning too soon afterwards. The teacher must allow time for consolidation of what has been read as a mental representation
  • practise inferential questions on aurally presented texts

With the first bullet point above in mind Kispal also reports that "the only condition that was found [by Hannon and Daneman (1998)] to significantly encourage inferencing was that of integrating questions into the text combined with allowing longer reading time" (this was in a study of university students rather than young children).

2. Teacher allows children to read the same part of the text

An end goal of reading instruction is to ensure that children can independently decode and understand something. Once the reading has been modelled it is a good idea for children then to have a go themselves in preparation for times when they won't have an adult to read for them. Typically we might ask children to do this in silence, but this isn't the only way. Re-reading aloud to a partner or to themselves has added benefits.

The Key Stage 2 Literacy Guidance Report from the EEF mentions that one way to improve fluency is for children to read aloud the same text that they have just had read to them. It also summarises research that shows that "fluent reading style supports comprehension because pupils’ limited cognitive resources are freed from focusing on word recognition and can be redirected towards comprehending the text." (page 11)

If re-reading a text develops fluency and fluency supports comprehension of the text then that is definitely something we should be building in to our reading lessons. This time spent re-reading also allows children to consolidate what they have heard and read (see Kispal's cautionary notes above).

3. Teacher provides a summary of the text

In his book 'Reading Comprehension: Assisting Children with Learning Difficulties' Gary Wooley outlines how mental models, or representations, are created by the reader:

"While reading, skilled readers normally develop a text-based model, which is a mental representation of the actual text discourse. The text-based model incorporates propositions extracted from the reading of successive sentences that are sometimes supplemented by inferences that are necessary to make the text more coherent."

I suggest that before a teacher models the inference-making that will lead to the creation of a more complex situation model (more on this in step 5) they should model a summary of the text to help children who have not developed a sufficient enough text-based model from which to begin to draw inferences. Providing summaries of the text for children is known to be a useful strategy to help EAL learners and so might they be for others learning reading comprehension strategies.

4. Teacher models inference-making

Kispal writes that "teacher modelling is regarded as the first step in training children to ask and answer questions of this type of themselves... Teachers should attempt to find texts rich in inferencing possibilities and to have in mind which inferences they will elicit in discussion."
(page 30)

The literature review then goes on to suggest that to show inference-making in use teachers should "model inferencing by asking relevant questions aloud and answering them" and that they should "think thoughts aloud to show how teacher arrives at an inference." 
(page 51)

Inference-making relies on the reader having done other things with the text such as clarifying word meanings, locating specific information and discussing necessary prior/background knowledge so these processes may need to be modelled also. When considering the activation of prior knowledge Kispal's review of research makes the following suggestions to take into consideration when discussing questions:
  • pupils generate initial associations 
  • they discuss and clarify their collective knowledge 
  • they reformulate knowledge, clarifying what they now know as a result of discussion

According to Kispal's review of literature, whilst modelling and discussing inference-making teachers should ask "questions about relationships between characters, goals and motivations" and ask "questions that foster comprehension monitoring, such as Is there information that doesn’t agree with what I already know? Are there any ideas that don’t fit together (because of contradictions, ambiguous referents, misleading topic shifts)? Is there any information missing or not clearly explained?" Teachers should always be asking "‘How do you know?’ whenever an inference is generated in discussion of a text." Teachers can also "show examples of how all types of questions can be derived from a text" using the question words (i.e. who, ‘when, why).
(page 38)

Questions that can be used to support systematic and structured teaching of the wide variety of inferences can be downloaded here: https://www.tes.com/teaching-resource/questions-to-support-inference-making-11825987

5. Teacher provides a second summary of the text which takes what has been inferred into consideration

In point 3 we looked at how teachers might share a summary of the text for the purpose of aiding the development of a text-based model. Once a text-based model has been created, and further inferences have been made, a situation model can then be developed.

In his book 'Reading Comprehension: Assisting Children with Learning Difficulties' Gary Wooley outlines how situation models (a kind of mental model or representation) are created by the reader:

"In contrast [to text-based models], situation models include elaborative inferences that integrate prior knowledge with text-based information.teacher modelling is regarded as the first step in training children to ask and answer questions of this type of themselves.

"Thus, the construction of a situation model is a dynamic constructive process that is determined by the interaction of the reader, the text structures, and the semantic content. 

"In constructing a situation model the reader is required to search for coherence at the local and global levels and to infer meanings that are often implied by drawing from their existing background knowledge. While doing this, the reader actively constructs the situation model by using information within the text and also information from stored prior knowledge. Thus, the main difference between text-based and the situation model is assumed to be one of inference making, the text-based model is inferentially light while the situation model is inferentially dense." 

It seems important to reassess the mental models that are created after making new inferences from the text.

6. Teacher reads aloud next part of the text

See point 1

7. Children read next part of the text themselves

See point 2

8. Children summarise text

See point 3. Children, having had this modelled to them, have a chance to practise their own summary to aid their text-based mental model before they answer any inference questions. This could be done in writing or verbally.

9. Children answer inference questions

This could be with support, without support, in pairs, independently, as a group, as a written task or as an oral task. Children may also need to understand the vocabulary used in the text, retrieve information from the text and link their background knowledge to the text - this could be done through discussion or by a structured sequence of questions (see my idea of scaffolding inference).

Kispal summarises that paired or group work allows pupils share the thought processes that led them to make inferences and that the younger the children, the more aural work they should undertake.
Kispal also writes that research on inference-making suggests that we should "train pupils to acquire the habit of asking themselves why-questions occasionally while they are reading, as these are most supportive of understanding". Another suggested strategy is to ask "pupils [to] generate questions using these question words [who, when, why etc] from a text and group members answer."
(page 38)

10. Children summarise text a second time taking into consideration what has been inferred

See point 5. Children, having had this modelled to them, have a chance to practise their own summary to aid their mental situation model once they have answered the inference questions. This could be done in writing or verbally.

Summary

Whilst structures like the one I've suggested can be useful, it is only there as a suggestion and will need to be adapted according to need. Having said that, this sequence takes into account many research-based practices which aid in the teaching of inference-making and therefore should be a good solid starting point for reading lessons that focus on inference-making (and probably other reading comprehension strategies). Use with discretion not because I said so!

For an example of how this might work with a real class novel, please see my planning for the first 10 chapters of 'My Dad's A Birdman' by David Almond. 5 whole lessons are focused on making inferences about characters' motives and a further 5 lessons focus on making inferences about characters' feelings. In the teachers notes I have not included information about the text summaries but every other part of the sequence is detailed.

Saturday, 27 January 2018

Research Findings: Are There Different Skills Within Inference?

In her literature review 'Effective Teaching of Inference Skills for Reading' (https://files.eric.ed.gov/fulltext/ED501868.pdf) Anne Kispal asks are there different skills within inference? and goes on to define and exemplify the most frequently cited inference types:

Coherence inferences maintain textual integrity. For example, in the sentence 'Peter begged his mother to let him go to the party', the reader would have to realise that the pronouns ‘his’ and ‘him’ refer to Peter to fully understand the meaning.

Elaborative inferences enrich the mental representation of the text, e.g: 'Katy dropped the vase. She ran for the dustpan and brush to sweep up the pieces'. The reader would have to draw upon life experience and general knowledge to realise that the vase broke to supply the connection between these sentences.

Local inferences create a coherent representation at the local level of sentences and paragraphs. This class of inferences includes:

  1. coherence inferences (described above).
  2. “case structure role assignments”, e.g. Dan stood his bike against the tree. The reader needs to realise that the tree is assigned to a location role.
  3. some “antecedent causal” inferences, e.g. He rushed off, leaving his bike unchained. The reader would need to infer that Dan was in a hurry and left his bicycle vulnerable to theft.

Global inferences create a coherent representation covering the whole text. The reader needs to infer overarching ideas about the theme, main point or moral of a text by drawing on local pieces of information (thus supporting my theory that one must be able to make inferences before trying to summarise a piece of text).

In 'Constructing Inferences During Narrative Text Comprehension' Graesser, Singer and Trabasso identify 13 classes of inference:

In all but two (or three) of the inferences in the right-hand column it is interesting to note that each class of inference contains a verb, and therefore is concerned with something being so. We might assume that most inferences are about action, state or occurrence.

The two (or three) classes of inference which appear not be concerned with something being so (or are not about action, state or occurrence) is Class 1: Referential, Class 8: Instantiation of Noun Category and potentially Class 5: Thematic.

These thirteen classes can be linked to Kispal's summary of the most frequently cited inference types:

"The order in which the inference classes are listed in Table 1 is not altogether arbitrary. Inference classes 1, 2, and 3 are needed to establish local coherence, whereas inference classes 3 and 4 are critical for establishing explanations. Classes 4,5, and 6 are important for establishing global coherence. Classes 7 through 11 are elaborative inferences that are not needed for establishing coherent explanatory meaning representations. Classes 12 and 13 address the pragmatic communicative exchange between reader and author." (Graesser, Singer and Trabasso, 1994)

Whilst the authors state that "these classes do not exhaust all of the potential inferences during comprehension" they provide a very good starting point to thinking about teaching inference-making at a primary level.

So, by and large, when we make inferences we are thinking about why things are or are not so.

This blog post is the second in a series of three:

Part 2: Reading Comprehension: A Structured Way Of Teaching Inference-Making
Part 3: Questions To Ask When Teaching Inference-Making

See also:

Scaffolding Inference - my blog post about how first asking vocabulary and retrieval questions can guide children towards answering inference questions.

Wednesday, 5 October 2016

Scaffolding Inference: Trialling a Teaching Technique

If you are short of time but would like to get the gist of this technique, please see my Quick Reference Guide: http://thatboycanteach.blogspot.co.uk/2016/11/scaffolding-inference-quick-reference.html


With inference being the most-assessed skill in the Key Stage 2 reading tests it is no wonder that teachers spend a lot of time attempting to teach children how to infer meaning from texts, with varying degrees of success. It's the sort of skill that readers (by that I mean those who make a regular habit of reading, and enjoy it) possess without really learning. Because of this, it is a skill that is hard to teach; many teachers infer naturally so deconstructing how they do it in order to teach a process to children can be difficult.

In case you missed it, the reading test framework has rearranged reading skills into eight content domains. The fourth domain, the one we are concerned with here, is: 
2d: make inferences from the text / explain and justify inferences with evidence from the text
Background Reading

The chapter in 'Reading Reconsidered' entitled 'Writing for Reading' (read an excerpt here) discusses the various structures a teacher might use within a reading session. The ideas presented widen the scope of how different task sequences can support the development of different skills. This made me think more carefully about how the teaching and learning sequence could build to help children to infer more successfully.

Penny Slater's helpful article 'Reading Re-envisaged' explores the links between vocabulary knowledge and inference skills initiated the thinking that led to my development and trial of this method. Her conceptual model (pictured left) represents how inference skills rely on good knowledge and understanding of vocabulary. In her own words: 
"...the model signifies the importance of vocabulary knowledge. If we consider each circle to be a moat which the children must cross before they are able to access the skills within the innermost circles, then we see clearly that they will not get very far if they do not understand the meanings on the words on the page. This chimes with what teachers are finding in their classrooms: lack of knowledge of vocabulary is a complete blocker. You can’t make any inroads into comprehension without addressing this issue first."
Anne Kispal's 'Effective Teaching of Inference Skills for Reading', in section 2.3 (page 26) goes into more detail on this and the document as a whole is an informative read. It has also been shown that 95%-98% of the vocabulary in a text needs to be understood in order to be able to derive a general meaning of the text (Schmitt, Jiang & Grabe, 2011).

So, another content domain comes into play, one which children must be confident with if they are going to be able to make inferences:
2a: give / explain the meaning of words in context
I also had an inkling that development of inference skills could be supported through the use of retrieval skills.
2b: retrieve and record information / identify key details from fiction and non-fiction
Children usually find retrieval easier than inference, however it is worth noting that in the 2016 KS2 tests even some of the retrieval questions were difficult, often because of the vocabulary skills that are needed in order to retrieve information. There are plenty of places to learn about how to improve vocabulary skills, so I won't go into detail on that in this article, but I must stress that it is important that children are taught skills such as contextual and morphemic analysis before they attempt the process I suggest. Before my own trial I spent around 4 weeks focusing on teaching vocabulary skills, allowing the children plenty of time to practice.

The Theory

The theory that I have been trialing is that inference skills can be taught by first studying the vocabulary used and then retrieving relevant information before going on to make inferences about a text. If inference is 'a conclusion reached on the basis of evidence and reasoning' then first a reader must be able to identify where the evidence is (retrieval) and before that the reader needs to understand the words used to present the evidence. In the model I propose (see right) the understanding of vocabulary is the foundation on which information retrieval is built, which in turn provides the support for making inferences.

The Practice

In short:
  1. Decide on an inference question (2d); the question stems based on the 2016 KS2 reading test made available by Herts for Learning on their blog are really useful for this.
  2. Begin to work backwards - work out where in the text the children need to go to locate useful evidence and ask a suitable retrieval question (2b).
  3. Continue to work backwards - which words or phrases do the children need to understand in order to be able to understand the evidence then ask a careful vocabulary question (2a).
  4. Once this process is complete (it may take a while at first), check that the 2a and 2b questions will adequately lead the children into answering the 2d question. If not, go back and tweak the questions.
There are different ways in which the 2a and 2b questions might provide a scaffold for answering the 2b question. In order to explain this I will share some examples. All the examples are based on 'Wonder' by R.J. Palacio. I chose 'Wonder' as our first class novel because although it is fairly heavy in subject matter, it is easy-going with its vocabulary. I wanted to begin by supporting children's acquisition of vocabulary skills in a non-threatening manner before we started to read novels with more advanced language.

The first excerpt takes place in the chapter entitled 'The Summer Table' in which a girl named Summer joins August who is alone at a lunch table on his first day at school.


In the first example (pictured above ) the scaffolding structure can be seen clearly: question 1 is a 2a question, question 2 is a 2b question and question 3 is a 2d question. There is a very obvious grammar discussion to be had to surrounding common nouns and proper nouns - the children asked for clarification on this despite the words in question 1 not being capitalised. The discussion we had cleared up possible later misconceptions that Summer meant the table was only for people named Summer - a misconception which would have been at odds with the basic fact that August was also sitting at the table. I've noticed that test questions are often set about texts with potential misconceptions so I try to take opportunities to incorporate similar tricky bits in my teaching.

The second example is taken from the same chapter; the text follows directly the previous excerpt:



The second example does not lead the children directly to the answer for question 3 but it does provide background knowledge which should inform their own thoughts on the motivation for Summer's actions. In answering question 1 the children realised that there was a long list of names and by answering question 2 they began to get the sense that the quote in question 3 was true; they gained their own insight into why August says that most of the names weren't actually summer names. Questions 1 and 2 allowed the children to understand what Summer was doing (making a long list, bending the rule that only children with summer names could sit on the table) before they began to think about why she was doing it.

Question 3 actually also requires previous knowledge of the text - the children must have already grasped that August (a boy with facial birth defects) is sitting alone on his first ever day in school whilst children whisper about his looks in order to infer that Summer agrees that so many children can sit with them so that he finds more friends. The more perceptive children might also realise that Summer also wants him to have fun so that he forgets about his situation and so that he feels like all the other children. I was satisfied that our previous reading and dialogic discussion (thanks Mat Tobin for the terminology) meant that they understood the whole text well enough to approach this question.

It should also be noted that here there are two retrieval questions and no vocabulary-based question; the vocabulary they needed had been covered in the previous set of questions.

Here is an example of a child's work. This task was undertaken independently directly after completing the previous task (see above). The first task was completed independently prior to a whole-class discussion and then children edited their answers (with a purple pen) based on the discussion that was had. This example contains no edits - the child was able to answer question 3 successfully first time. It is worth noting that this child is one of the best readers in my class - for her the scaffold has had almost immediate impact. In further blog posts on this subject I will provide before and after evidence.


For the next examples I must give credit to Rhoda Wilson for her excellent 'Moving Beyond Comprehension Sheets' resource as I used it along with the Herts for Learning question stems to vary the question styles in these activities.

Here's an example of a very scaffolded set of questions - the scaffold questions (questions 1, 2 and 3) make the answer to question 4 very obvious.


This one worked so successfully that I actually encouraged the children to further their answers for number 4 by explaining how the evidence showed that the children were unsure how to treat August - this was not initially required of them, and when compared to similar questions in the 2016 KS2 test, this would be classed as an inference (2d) question without the addition of an explanation. It also made me contemplate giving them the inference question to answer before the scaffold questions, as well as after, in order to compare the difference and the impact the scaffold questions have on the quality of answer.

Some more activity examples:


Here is an example of child's work. This child entered year 6 in September assessed at a year 4 standard for reading. This method appears to have been very successful for him, even after only a few times working in this way.


One more example:
What Next?

If this way of scaffolding inference questions works for the children in my class then I will begin to adapt it in order to support the development of skills outlined in the other content domains:
2c; summarise main ideas from more than one paragraph
2e: predict what might happen from details stated and implied
2f: identify / explain how information / narrative content is related and contributes to meaning as a whole
2g: identify / explain how meaning is enhanced through choice of words and phrases
2h: make comparisons within the text
It will also be important to begin to remove the scaffolding - for some children sooner than others - in order to encourage children to use the skills independently; one question often raised against methods such as this is how will this approach help children when the structure is removed, for example, in the SATs reading test? And it's a good question. My hope is that it will provide them with a method for answering inference questions; a method which will be embedded in their way of working. If this technique is successful then children will naturally make inferences using their ability to understand the vocabulary (these skills will need to be taught in addition to this method of scaffolding questions) and their ability to locate and retrieve information from the text.

The trial of this technique for scaffolding inference is in its infancy. As such I will follow up this blog post with others including commentary on what I learn, further examples of questions and some more examples of children's work showing the impact.

I would also love to engage in discussion on this idea - please use the comments section to tell me where I am going wrong, to point me in the direction of relevant research or additional reading or to share your own examples if you decide to try it!

Click here to read a testimonial from one teacher who used the technique.

Click here to read about how this, and other changes made to the way we teach reading, impacted on our SATs results.

Saturday, 27 January 2018

Questions To Ask When Teaching Inference-Making

The EEF KS2 Literacy Guidance document states that "reading comprehension can be improved by teaching pupils specific strategies that they can apply both to monitor and overcome barriers to comprehension". It goes on to say "strategies should be modelled and practised to ensure they become embedded and fluent". It concludes that "The potential impact of these approaches is very high, but can be hard to achieve, since pupils are required to take greater responsibility for their own learning. This requires them to learn three things: what the strategy is, how the strategy is used, and why and when to use the strategy. Developing each of the strategies requires explicit instruction and extensive practice".

In order for children to make inferences independently the EEF's gradual release of responsibility model is useful. It describes how greater responsibility for using these strategies can be transferred to the pupil:

1. an explicit description of the strategy and when and how it should be used;
2. modelling of the strategy in action by teachers and/or pupils;
3. collaborative use of the strategy in action;
4. guided practice using the strategy with gradual release of responsibility; and
5. independent use of the strategy.

In my last blog post 'Reading Comprehension: A Structured Way Of Teaching Inference-Making' I concluded that children will probably benefit best from having the chance to practise specific inferences based on the different types of inference-making listed above. If teachers can provide questions that are of a similar structure, and provide structures for the answers too, then children who are at the learning stage of inference-making might have a better of chance of being able to make inferences whenever they are reading.

With this in mind, here are some inference question focuses that might help teachers to structure their lessons and questioning more carefully in a way that allows them to model particular skills which the children can then practise:

Making inferences about actions

Ask questions about:
  • how a character feels
  • why a character feels a particular way
  • why a character acts/behaves in a certain way (motives)
  • why a character says certain things (motives)
  • why a character says things in a certain way (motives)
  • why a character does things in a certain way (motives)
  • what a character thinks
  • why a character thinks/believes/expects (etc) certain things
Questions may be framed in many different ways. Here are some examples, given in an order that might allow gradual transfer of responsibility to children. I suggest this sequence of questions takes place over a series of lessons, rather than in just one lesson, especially where written answers are required:

<quote from text> This tells is that x feels... <multiple choice answers> 

Provide a quote from the text that children can infer information from, provide a description of what is felt/said/done and then give a choice of possible inferences for children to choose from. Questions like this might be used more often with younger children. As exemplified in the 2017 KS2 Reading SAT below:


Character feels x. How do you know? 

Possible answer structure: In the text it says ________________. This shows the character feels x because _______________________.

Give a description of what is felt and ask for children to locate information that supports this theory. Children will probably benefit from being asked several questions with the exact same question and answer structure in order to practise based on what the teacher has modelled, as exemplified here:



This type of question is also exemplified in the 2017 KS2 Reading SAT below, where the cat is the character and the lack of enjoyment is the feeling and the request for three ways in which the cat shows this is in place of the How do you know? question:


How does x feel? What does x think? Explain your reasons. 

Possible answer structure: X feels ________________. I know this because in the text it says ________________. This shows the character thinks x because _______________________.

This is more difficult than the previous questions as the child has to do more: they have to find their own word to describe what is felt/thought etc and they have to support it with evidence from the text. Ensure that children are presented with several opportunities within a lesson to answer questions at this difficulty level - keep the lesson focused on this one question type. As exemplified in the 2017 KS2 Reading SAT below where the question asks about thoughts of expectation:


Making inferences about events

Ask questions about:
  • what happened (where details are not given explicitly and retrieval skills can't be used)
  • why something happened
  • where something happened
  • when something happened
  • how something happened
  • why something happened in a certain way
  • what was unusual or different about what happened
  • how something has come to be
Questions may be framed in many different ways. Here are some examples, given in an order that might allow gradual transfer of responsibility to children:

Questions about what happened:

<quote from text> What do you think happened? <multiple choice answers> 

What evidence is there that x happened?

Possible answer structure: In the text it says ________________. This shows that x happened because _______________________.

In the paragraph beginning... what do you think happened? Find two pieces of evidence from the text to support your answer.

Possible answer structure: I think that _____________________ happened. I know this because in the text it says ________________. This shows that x happened because ______________________.

Questions about why something happened:

Read the story/paragraph beginning... Join each event to its cause. <provide a list of events next to a jumbled list of the causes for each event>

<quote from text> Why did x happen? <multiple choice answers> 

x happened. Why did this happen? Give evidence from the passage.

Possible answer structure: I think x happened because in the text it says ______________________. This suggests that __________________________________.

Questions about where something happened:

Read the story/paragraph beginning... Join each event to its location. <provide a list of events next to a jumbled list of the locations of each event>

Where did x take place? <multiple choice answers> 

Where was a when x happened? Explain how you know using evidence from the text.

Possible answer structure: I think a was _________________________. I think this because in the text it says ______________________. This suggests that __________________________________.

Making inferences about state

Ask questions about:
  • what something is
  • what a place or object is like
  • why a place or object is as it is
  • what we know about someone's character (what a person is like)
  • where something is (different to where something happened)
  • why something is where it is 
Questions may be framed in many different ways. Here are some examples, given in an order that might allow gradual transfer of responsibility to children:

Questions about what something is:

Tick two pieces of evidence from the text that tell us that the object is x. <provide several quotes from the text which may or may not provide evidence for the state of x>

The object is x. Find three pieces of evidence from the text that support this theory.

<provide an excerpt from the text> What does this suggest that x is? Give your reasons.


Possible answer structure: This suggests that x is ________________. I think this because in the text it says ____________________________.

What is x? How do you know?

Possible answer structure: x is _______________________. I think this because in the text it says ______________________.

Questions about what a place is like:

The place is x. Which of the options below could be used as evidence? <provide several quotes from the text which may or may not provide evidence for x>

The place is like x. Find supporting evidence in the text.

Possible answer structure: This suggests that x is ________________. I think this because in the text it says ____________________________.

What feeling does the character get of place x? What evidence is there in the text?

Possible answer structure: The character thinks the place is ______________________. I think this because in the text it says ____________________________.

Questions about what a person is like:

<quote from text> The character is made to seem...: <multiple choice answers> 

<quote from text> How is the character made to seem x? Explain two ways, giving evidence from the text to support your answer.

Possible answer structure: In the text it says _____________________ which suggests that the character is _________________________.

As exemplified in the 2017 KS2 Reading SAT below, where the character is the whale and the characteristic (mysterious) is given:

<provide an excerpt from the text> What impression does this give us of the character? Give your reasons.

Possible answer structure: In the text it says _____________________ which suggests that the character is _________________________. It also says that __________________________ which makes the character seem ______________________.

As exemplified in the 2017 KS2 Reading SAT below where the characteristic is not given and information to support the answer has to be found:

Read the whole text. Which aspects of the character's personality change? Use examples from the text.

Possible answer structure: At the beginning the character is _______________ but by the end they are ________________________. I know this because at the start the text says ________________________ and at the end it says __________________________.

This is an example of making a global inference based on understanding of the whole text. This can be more difficult than making local inferences about small parts of the text (as in the previous question examples).

A note on answer structures: the examples given are all full sentence answers - you may want to teach ways of being more concise in order to save time, particularly with timed-tests in mind. Bullet-pointing and more note-like answers are often good for this.

These questions and more can be downloaded as a simplified word document at TES resources.

I have by no means covered all the possible kinds of inferences in this blog post, nor have I exemplified them all. Hopefully what I have managed to convey is:
  • Inference-making can be modelled by the teacher and practised by the children
  • Teachers can ask specific kinds of questions to provide practise of inference-making
  • Children can practise specific kinds of inference-making
  • Children can be provided with structures to help them answer questions
  • There are a levels of question difficulty within each kind of inference questions
  • Children can be given multiple opportunities to practise each kind of question, especially where there is a written answer
For an example of how this might work with a real class novel, please see my planning for the first 10 chapters of 'My Dad's A Birdman' by David Almond. 5 whole lessons are focused on making inferences about characters' motives and a further 5 lessons focus on making inferences about characters' feelings. The first lessons in each section feature multiple choice questions, moving onto questions which require increasingly more writing using an answer structure.

This blog post is the fourth in a series of four:


See also:

Scaffolding Inference - my blog post about how first asking vocabulary and retrieval questions can guide children towards answering inference questions.
How To Write Good Comprehension Questions - this blog post goes into more detail on what else to take into consideration when it comes to writing your own comprehension questions.

Wednesday, 9 November 2016

Scaffolding Inference (Quick Reference Guide)

Inference skills in the new cognitive domains are summarised as:

2d: make inferences from the text / explain and justify inferences with evidence from the text

Penny Slater's helpful article 'Reading Re-envisaged' explores the links between vocabulary knowledge and inference skills. Her conceptual model (pictured left) represents how inference skills rely on good knowledge and understanding of vocabulary. In her own words:

"...the model signifies the importance of vocabulary knowledge. If we consider each circle to be a moat which the children must cross before they are able to access the skills within the innermost circles, then we see clearly that they will not get very far if they do not understand the meanings on the words on the page. This chimes with what teachers are finding in their classrooms: lack of knowledge of vocabulary is a complete blocker. You can’t make any inroads into comprehension without addressing this issue first."

So, another cognitive domain comes into play, one which children must be confident with if they are going to be able to make inferences:

2a: give / explain the meaning of words in context

This approach also explore the possibility that development of inference skills could be supported through the use of retrieval skills.

2b: retrieve and record information / identify key details from fiction and non-fiction

The Theory

The theory that I have been trialing is that inference skills can be taught by first studying the vocabulary used and then retrieving relevant information before going on to make inferences about a text. If inference is 'a conclusion reached on the basis of evidence and reasoning' then first a reader must be able to identify where the evidence is (retrieval) and before that the reader needs to understand the words used to present the evidence. In the model I propose (see right) the understanding of vocabulary is the foundation on which information retrieval is built, which in turn provides the support for making inferences.

The Practice


1. Decide on an inference question (2d); the question stems based on the 2016 KS2 reading test made available by Herts for Learning on their blog are really useful for this.
2. Begin to work backwards - work out where in the text the children need to go to locate useful evidence and ask a suitable retrieval question (2b).
3. Continue to work backwards - which words or phrases do the children need to understand in order to be able to understand the evidence then ask a careful vocabulary question (2a).
4. Once this process is complete (it may take a while at first), check that the 2a and 2b questions will adequately lead the children into answering the 2d question. If not, go back and tweak the questions.

For a more in-depth exploration of this technique, including examples of questions: http://thatboycanteach.blogspot.co.uk/2016/10/scaffolding-inference-trialling.html
Many of the examples from the blogpost are available for download here:
https://www.tes.com/teaching-resource/-11416437

Tuesday, 17 October 2017

How To Write Good Comprehension Questions

Many of the commercially available reading tests often have some questions which are poorly written, so poor that it is unclear as to what the questioner is getting at. These ambiguously written questions can hardly help children to develop their ability to understand what they have read, nor can they be a useful assessment of what they understand. The problem is that asking good comprehension questions that make sense to children is really difficult – it’s an art we have to practice, and that we might not always get right. But what can we do to ensure that we provide the best questions possible in our reading lessons?

Choose Texts Carefully

Before thinking about the questions, you need to select the text carefully. If you know that you want to practise a particular reading skill then the text you choose needs to support that. For example, there would be very little point in choosing an instructional text to teach and practise inference skills. That is an extreme example, but it makes the point - make sure the text you choose supports the skills you want to teach.

Of course, sometimes you might select a text and the focus of that lesson will be simply understanding the text as a whole - in a lesson like this you wouldn't want to focus solely on teaching one particular skill, you'd want to ask the necessary questions which really ensure that children are reading for meaning.

Plan Ahead

Even the most experienced teachers run the risk of asking superficial and poorly-worded questions if they have not pre-read their text and planned out the questions they are going to ask. The tendency also is to ask low-level literal questions (retrieval) rather than any other kinds of questions which probe deeper into a text. Write down the questions you want to ask.

At this point I should point out that when I refer to writing good comprehension questions I don't necessarily mean questions that will be presented to children in written form. The questions that you write might only go as far as your planning sheet - in an actual lesson they will be questions that you pose orally. The same goes for the answers that children might give - they could be written or oral.

Use SATs Question Stems

One thing I would have to say in favour of the SATs is that at least they are well written and to a proficient reader the answers, however difficult, are not ambiguous. Sometimes perhaps the mark schemes are a little narrow, and children don’t get marks when they clearly have shown understanding, but that is not to do with the way the questions are asked.

We can learn how to write better questions by studying the KS1 and KS2 tests but unfortunately not all teacher in all year groups are familiar enough with them. Several useful documents have been produced containing the question stems from the most recent tests:
The questions in the documents above are organised into the different content domains – for an easy way to remember these content domains, please see my blog on the Reading Roles.

Use Different Response Formats

Many comprehension activities set by teachers are in the simple form of a written question. A quick flick through a test will provide plenty of other ideas for how to present questions:
  • Multiple Choice (tick or circle)
  • Ordering Events by Numbering
  • True or False
  • Matching
Some examples (the colours and symbols here relate to the Reading Roles):





It is well worth creating a word document that contains some pre-made questions like this to copy and paste each time you create a reading comprehension activity. I have plenty of examples available on TES so you don’t even need to create them yourself.

Although not featured recently in tests, not in their strictest form anyway, cloze tasks are a good way to test reading comprehension. When creating a cloze procedure it is best to remove words that are crucial to the meaning of the original text that children read. Cloze procedures can also include a multiple choice element if the line for the missing word is followed by a choice of several words for children to choose from.

In their book Understanding and Teaching Reading Comprehension, Jane Oakhill, Kate Cain and Carsten Elbro recommend true/false, multiple choice and cloze task questioning formats, but point out that the different formats lend themselves better to differen question types. For example, true/false judgements are better for retrieval questions, whereas a multiple choice question might help children to recognise an inference that can be made from what they have read.

Research shows that presenting questions in a true/false format is also good for children for whom English is an additional language. Click here to read more on this.

Focus on a Particular Skill

The temptation is to just ask the first questions that come into one’s head when reading the text intended for the reading lesson. This is fine for a summative assessment (that mixture of question types and skills is how the SATs are presented) and for understanding a text, but isn’t great for teaching children specific comprehension skills. If children are only ever presented with a scattering of questions across the content domains there is little opportunity for deliberate teaching and practice of particular skills such as inferring, summarising or predicting.

Most lessons should focus on one skill; sometimes a whole sequence of lessons will be focused on the teaching and practise of one skill. I have written more about how to break down the different kinds of inference so that they can be focused on in lessons: Questions To Ask When Teaching Inference-Making

Scaffold Answers

There are several ways to do this so that you support children. I have written a lot about scaffolding inference by first asking vocabulary and retrieval questions, but there are other scaffolding structures, too.

Here's an example of scaffolding inference:


Before being able to make plausible predictions children might first need to answer relevant questions about vocabulary, they will then need to be able to retrieve information and make inferences based on them – prediction is a form of inference.

In order for children to answer questions about author’s purpose they might need to first answer well-crafted questions about vocabulary, retrieval and/or inference.

Similarly, in order to summarise information children may need to go through the process of answering relevant questions about vocabulary, retrieval, inference, author’s purpose before they can give an accurate summary of a text.

When writing sequences of questions like this it is a good idea to start with the final question you want children to answer, and to work backwards from there – the children should be able to draw on all the information they have given in previous questions to answer the final question.

To read more about potential scaffolding structures click here: http://www.thatboycanteach.co.uk/2017/12/scaffolding-structures-for-reading.html

Give Relevant Information

Don’t leave children searching forever for the place in the text where they might find their answer. Give them pointers such as:
  • Look at this sentence: 
  • Look at the paragraph beginning 
  • At the top of page it says…
Some examples:



Even the tests provide this sort of information. Children cannot demonstrate their comprehension skills any better without this information, although by giving no such clues children may practise their scanning skills.


Checking Your Questions

Once you have written your questions it is a good idea to either get another teacher to have a look through them, or to return to them later and read them with fresh eyes: Do they make sense? Is it clear what the answers should be? Do they need re-wording? Ask yourself these kinds of questions and edit accordingly - you don't want children to be turned off answering the questions due to a lack of clarity.

As Oakhill et al point out in 'Understanding and Teaching Reading Comprehension', it is important to check that the questions you have written actually check comprehension of the text. They outline how some questions could be answered using prior knowledge solely, without reference to the text at all.

Pie Corbett made a few salient points to me on this issue which I'll use to conclude:

"You have to be able to find a text worth reading then design questions (or focuses) that challenge and begin to deepen thinking. It is worth thinking about questions that are worth asking and ones that are not worth asking."

Monday, 14 May 2018

Teaching Reading Comprehension: Modelling and Practice (Example Lessons)

I’ve been thinking, reading and writing a lot lately about how we teach reading strategies and skills in primary schools. I won’t bore you with all the details but thought I’d simply share some lessons that I’ve prepared for some year 3 teachers at the schools I work in. If you want to find out more about what I’ve been discovering, and the thinking behind the lessons I’ve planned, I’ve provided some links at the end of this blog post.

These lessons, although not fully-formed (I didn’t want to dictate everything), are a good representation of how I think teachers should model the use of reading strategies and skills in a lesson and how children can be given practice of using the same strategies and skills that their teachers have modelled. The lessons involve both opportunities for oral and written comprehension activities; the written activity can just as well be worked on orally, although it is designed so that children can work on it independently by giving written answers.

Some of the lessons you will see here were based on versions of Aesop's Fables written by Michael Morpurgo and illustrated by Emma Chichester Clark, published by Orchard Books. Although the book hadn't been chosen with inference-making in mind, it was serendipitous that there were plenty of opportunities to focus a few lessons on that particular skill. Hopefully these examples will show that, even whilst having a focus on a particular strategy or skill, other strategies and skills might be used in support whilst developing the skill which is the focus of the lesson (in this case inference-making).

For each lesson I outlined the L.O. (based on the National Curriculum POS for year 3/4 in this instance) and some introductory questions and items for discussion:



I then suggested some exemplar questions for the teacher to model which focus on the lesson's L.O.:

All of the above could be done as a whole class reading lesson, or as a guided group. The point of all of the above is to have discussions about the text and to orally develop strategies such as clarifying (what do the words mean?) and inferencing (why do the characters do what they do?). The intention is that children will then be better prepared to have a go at some similar questions themselves without the teacher having already answered them by way of demonstration.

In this particular example the questions are focused around multiple choice answers with the hope that children will consider each option in order to decide whether or not it is good evidence for the character's motives. Notice that not all the questions are inference questions; other questions are asked which might support the child's understanding so that they are able to make the more difficult inferences (see my blog posts on scaffolding for more information on this idea).




For more information on the symbols/colours use in this example, please read the following: http://www.thatboycanteach.co.uk/2016/12/reading-roles-cognitive-domains-made.html

This part of the lesson could be done as an independent written activity or as part of a guided group. The multiple choice questions should spark some good discussion about why the correct one is correct and about the reasons children have for selecting their answers. If this was being completed as an independent written task there is the potential for a follow-up written task asking children to give their reasons for their selection.

Following this, and in order to practice another strategy, I suggested the following:

The following lesson follows a similar structure:




You can download these resources on TES - they are editable so even if you don't have the book, you can use the activities as a template: https://www.tes.com/teaching-resource/aesop-s-fables-reading-comprehension-teacher-notes-and-pupil-activities-11900274

These two lessons represent the first two in a potential sequence where children might move beyond being given multiple choice options. In another sequence of lessons based on David Almond's 'My Dad's A Birdman' children moved onto giving spoken and written answers to inference questions (which throughout the sequence all focused on characters' actions only). To begin with they answered questions with a structure that had been provided and modelled to them, as exemplified in the teacher notes:


They then answered their own questions. Again, this could be done independently, collaboratively or as part of a guided group with a teacher:

The children spent two lessons practising this before being shown how to further add to their answer, as demonstrated in the teacher notes:


The children then practised using this addition to the answer structure (although they only practised one as this was a chance for teachers to assess children's attempts at what is quite an advanced skill for year 3 children):

In the sequence of lessons on My Dad's a Birdman children spent 5 sessions focusing just on making inferences about character's actions followed by another 5 sessions focusing on making inferences about characters' feelings. For more on why there was such a sustained focus please read my blog post entitled 'Reading Comprehension: A Structured Way Of Teaching Inference-Making'. Along the way the children also exercised other reading comprehension strategies and skills in order to support their inference making and general understanding of the text. They also spent time just reading the book and enjoying - teachers and children alike kept telling me how much they loved the book. The fact that they had spent time completing such activities as outlined above enabled them to enjoy the book, rather than spoiling their enjoyment of it.
See my blog post entitled 'Giving the Gift of Reading: Activities That Promote Reading for Pleasure' for more on this.

You can download these resources on TES - they are also editable so even if you don't have the book, you can use the activities as a template: https://www.tes.com/teaching-resource/inference-questions-for-my-dad-s-a-birdman-by-david-almond-chapters-1-to-10-inc-teacher-notes-11842172

Further reading from my blog on teaching reading in primary schools:

Teaching Reading: A Simple Approach
Reading Roles: Elements Of The Content Domain Made Memorable
Reading Roles PLUS: Teaching Metacognitive Reading Comprehension Strategies (not exemplified in this blog post)
Questions To Ask When Teaching Inference-Making
Should We Teach Reading Strategies In Isolation Or In Combination?