David Francis

DAVID FRANCIS

Sure.  I’m David Francis and I’m a professor of psychology at Quantitative Psychology at the University of Houston, and the director of the Texas Institute for measurement, evaluation, and statistics.

To understand research teachers need to – to know that uh science is – science knowledge is different from typical every day knowledge in that scientists are trying to make inferences about relationships among things where those inferences are always based on a certain amount of uncertainty and when we as scientists evaluate research we – what we know about something in science is – is based on not a single study, but on a collection of studies.  And so teachers need to know that not a single study will tell us the relationship between things but will point us in a direction.  And the value that any particular study has is the degree to which is can reduce uncertainty about the relationship between things.  And so different studies have different value in the sense that they uh have a different degree to which they will reduce uncertainty about a phenomenon, but – but in the end it’s always the body of evidence around a – relationships between things that tells us whether or not those things are connected in some causal – causal way.  So teachers need to understand that when they’re looking at research to not look just at a single study, but to look at a collection of studies that deal with a phenomenon, that that’s where the knowledge comes from. 

Teachers – when teachers think about research they have to question uh the methodology that was used and whether the methodology maps well to the question that was framed around the research.  Different methodologies are better equipped to answer certain kinds of questions then other methodologies.  So, for example, quantitative methodologies are better equipped for addressing issues of causality.  Does ‘a’ cause ‘b’.  Does doing this kind of instruction lead to these kinds of gains uh in – in student performance?  Qualitative research is better equipped to describing phenomena.  What does it mean to provide this kind of instruction?  What does this kind of instruction look like?  If I’m a teacher that’s providing um primary language support for second language learners in their literacy instruction, what does that look like?  What does it mean?  So that’s more easily captured through qualitative research methodologies then through quantitative research methodologies.  So teachers need to understand that different methodologies are better equipped to answering different kinds of questions.

The challenge here is reframing these questions in a way that – um there’s a really interesting history around the use of discrepancy as a means for identifying children with learning disabilities and in particular it – the use of discrepancy as a method for identifying children really stems from states requests to the Department of Education for assistance in the process of identifying children.  When the law was first written uh to – to define um disabilities there is no mention in the original law about discrepancy.  There’s mention only about impairment and impairments in various things.  But when the regulations were written around that law to try to help states actually operationalize the law, that’s where discrepancy crept it and it really stems from – from conceptualizations of disabilities as unexpected under achievements.  So students are achieving at a level that is below what’s expected for them, and so that – that created the need to define “Well, how do we determine the expectation?”  And expectations were one – one way – easy way to start thinking about that is in terms of – of uh what the student’s intellectual potential would be and that then gets operationalized in terms of an IQ test.  But uh I think current thinking and current work is actually pushing away from the use of the discrepancy model because we feel that it’s – it’s really not getting us where we need to go.   

I think that what teachers need to understand about discrepancy in terms of how it fits into uh making placements is that it – it really doesn’t have a strong role there.  It is really an eligibility issue and one of the things that uh – that people within the field – and I think also to – to a certain extent advocacy groups are pushing for – is that we focus more on providing students with what it is that they need to be successful learners and less on quala – qualifying that they’re eligible for certain kinds of supplementary services.  So trying to shift the burden of effort towards providing the interventions and the services that students need and less on making sure that we’ve uh, uh determined the eligibility criteria and that uh – that students are in fact eligible for the services that – that they’re getting.  So I think what teachers need to understand about discrepancy in terms of placement is that it is just a means to an end and the goal is really to see that the students are getting the services that they need.

Yeah, I think that’s the current thinking and to try to move more towards a um – a model that – the model that certain kinds of instruction need to look like to be effective and if students are getting that kind of instruction and then you’re modifying it based on the fact that students are making progress and then you have a – an extended history of the right kind of instruction that’s not leading to academic outcomes that are desired, then you have evidence that there’s a disability and at that point you can make the determination and say, “There’s – there’s a disability” as opposed to making the determination there’s a disability and then providing the instruction.  So let’s provide the instruction that kids need and if we have to attach a label to it we can do that after the fact.  So…

My feeling about the labeling is that the labeling is uh – unless the label provides some help in working with the student then it’s not serving a useful purpose and so I would want to avoid the use of the label until it conveys some knowledge about what to do and uh – and how to work with that particular student.  So the labels have to carry some value added beyond just saying that it makes the student eligible for certain kinds of services and – and – you know, for me.  But – but again, bear in mind I’m a quantitative psychologist working inside a – inside an academic department.  So…

Okay, uh the connection between spelling, phonetic instruction, and reading, we have to distinguish between phonetic based languages and non-phonetically based languages uh in terms of phonetic – phonetic writing systems and non phonetic writing systems.  But these uh systems are actually very tightly integrated and they are highly uh connected in terms of – when students learn to read or when individuals learn to read in a phonetically uh transcribed – phonetically written language, they have to learn how the sounds of the language are mapped to print.  Well that mapping process is a – is a process of using letters.  So the phonetic uh relationship there is understanding these connections between sounds and print, and of course, spelling is the – the regurgitation of those sound, print, uh relationships – or the print relationships, but certainly it maps in – you know, internally to some phonetic for the lan – for – for the words and that’s in fact what – what students are doing in the process of reading is in attempt to access the meaning, generally speaking, for most words they’re accessing that meaning through the phonetic representation of that word.  So we learn language as a sound-based system.  So our connections to the meanings of words is through that sound-based system.  So when we’re – when we’re actually extracting meaning from print it’s mediated by that – by that sound relationship.  So these systems are actually very tightly integrated and you’ll find that when you look at – at spelling and decoding skill, the ability for students to read uh, uh nonsense words, for example, those are very tightly connected.  That students who are better spellers are better equipped, better able to read nonsense words.  They just do a better job of that.  So I think in – in alphabetic languages there’s a really tight connection between spelling, and reading, and phonetic knowledge. 

Uh, a phonological impairment is an inability to work within the phonetic language, and in a sense it’s an inability to manipulate the – uh the sound parts of language and in – it’s – it has to do with these connections between sound and print is – is where the disability – where it impacts on students’ inability to learn to read.  One of the things that we know is that most students who struggle to learn to read have problems at the word level, so they struggle at the decoding process.  That is that process of looking at a word and determining what is that word and so – and it turns out that when you do assessments with these students and you do assessments that require them to listen to a sound and then manipulate that sound, or to listen to isolated sounds and put them together, or to listen to a word and then break it up into sounds, they really struggle with those kinds of tasks.  Students that struggle to learn to read are really poor at playing Pig Latin because they take that word and strip off the initial sound and attach it to the back.  And so one of the things that we see empirically is that students who struggle to learn to read have a lot of problems with this manipulation of speech in terms of sound.  That is the internal elements of words in terms of sounds and manipulating those.  and it’s – you know, there are a number of different kinds of studies that look at this both empirical in terms of doing assessments with kids, but also uh functional imaging studies that actually look at how kids brains function when they’re processing text and when they’re processing words and show – those studies show that there are actually differences in terms of how kids are uh – in terms of their neuroanatomy, how they’re actually functionally approaching those problems.  Now the interesting thing is that those studies also show that uh when you work with students and you get them the right kind of instruction that the – those patterns of functional organization in the brain change and they start looking more like what typical readers look like.  So it’s a software problem; it’s not a hardware problem.  And so it’s – it’s a matter of learning how to work with the sounds and the speech stream and manipulate them and the students – some students need additional instruction in terms of how to be able to do that.

When we think about development we always – uh, we try to think in terms of um – across a broad span of time and when we think about developmental processes and early pro – and early developmental process and a late developmental process, uh those can be 2 different processes that are simply connected, one causing the other, or it could be some kind of an unfolding phenomenon where this proc – the process – it’s one process but it looks different early then the way it looks late.  And those are 2 different kinds of developmental phenomena both of which can be studied using quantitative methodologies.  We also have to distinguish between whether or not the – the – the developmental process is actually changing qualitatively even though it’s a – a singular process let’s say.  For example, uh people might think in terms of – of (unintelligible) stages of cognitive development that we might think about that as a – a singular entity even though qualitatively it looks different at different stages.  So when you want to know about how quantitative research methods can be used to study an early process and its influence on a later process, we have to think about “What are the nature of those processes?”  Are we talking about a qualitative change in the process?  Is it a quantitative change where it’s just – something is changing in terms of its level of expression, um and is it 1 process or is it 2 different processes, one of which impacts the other?  So once you know those things, once you’ve sort of articulated what it is you’re trying to study, then the methods are fairly straight forward in terms of how to link them up.  But the – the fundamental thing that has to be there is longitudinal data.  That is data on the same units from early into late.  You have to study the same individuals across that developmental continuum, otherwise there’s no way to actually do that.  We can’t study those kinds of developmental influences by studying different individuals at one age and another set of individuals at – at a later age.  That won’t help us to understand how this process informs the other process.

So the question is “Has research shown that – that development is both additive and in some cases discontinuous?”  And I think – I try to think about examples of where that might be, uh and I believe that there are – there are examples of both kinds of – of – of development.  Um when we look at reading as a process, uh there are aspects of reading which are quite additive in nature.  For example, uh children learn first – in reading an alphabetic language they learn first the letters.  They learn the names of those letters, the sounds that they make.  They learn how to put them together to form simple words uh and how to manipulate – uh manipulate sounds, uh, though engaged in this process of decoding and understanding.  And then as they get better at that process becomes more and more fluent and uh more automated so that they can actually execute that (unintelligible) without any conscious effort um all of which leads them to a stage where there are – where they are better equipped to understand what it is that they’re reading because less and less of their effort is going into the end coding or decoding process and more and more of it can be focused on what has been decoded and – and how those things are connected up to each other.  So when you look at reading as a process it is in some ways discontinuous, but also in some ways quite continuous in that there’s – there’s additive changes or quantitative changes that are happening.  You can’t be a fluent decoder before you are a decoder.  You – you won’t understand the print until you’re able to decode it.  So these things build on one another and yet they’re quite distinct in terms of their – their nature.  The – the – the process of distracting meaning is quite different from the process of decoding, although the one builds on the other.

There was some writings uh that took place I think in the early 70’s, maybe late 70’s around this whole issue, the idea that uh – about lag versus deficit and do children with disabilities simply lag behind their age development peers – their age peers or do they actually show a deficit relative to those students?  And just at the most basic level this is important because if it’s a lag issue then we might say, “Well, it’s maturational.  If we give students time they’ll catch up and it’s just a matter of – of sort of meeting them at their current needs and making sure we continue to – to monitor their development.  But basically everything’s okay they’re just running a little slow.  Um, surprisingly there’s been very little work to explicitly test this idea, um, but we actually did do some work and this is one of the questions you were going to ask me.  This is one where actually recite some of the stuff we’ve done.  We actually did a study um, um – uh Sally Shaywitts and Venit Shaywitts at uh – at Yale University – Sally was the principal investigator of a study called “The Connecticut Longitudinal Study.”  And this was an epidemiologic study of a large cohort of kindergarten children in the state of Connecticut um and they actually have now studied – these children were recruited in kindergarten and then they were followed uh each year all through high school and actually on into adulthood.  And at the time that we embarked on this particular assessment of this idea of lag versus deficit we first studied the data through 6th grade, and then later through 9th grade, and then later through 12th grade, and actually tested this hypothesis multiple times to see if in fact it was just a problem of not having had enough data.  But there were about 400 students in this uh investigation and I actually used models uh for change in reading to actually model the reading data longitudinally and compared students who had not been identified as having a reading disability versus students who were identified either through a discrepancy model of a difference between IQ and achievement or were simply identified as being low achieving.  And those identifications were done at 3rd grade.  But then we modeled the data from 1st grade on through either 6th grade, 9th grade, or 12th grade depending on which – which version of the study we were looking at.  And the data uh quite clearly indicated that uh it – for this data set those students identified as being disabled clearly were at a deficit relative to peers and they – that – that deficit did not diminish over time.  There is a very, very slight diminish of it but it’s almost imperceptible, you know, in terms of the magnitude of the differences uh over time.  And what’s also equally interesting is that there is virtually no difference between the students identified as disabled because of discrepancy versus those identified as disabled because of low achievement in 3rd grade.  Their growth parameters, their growth trajectory are literally right on top of one another.  And so this is important because it says that unless we do something, unless we intervene to change the trajectory that the student is on they will end up at a lower level of performance then their non-disabled uh, uh peers.  And so this – it’s – and I’m not aware of anybody having attempted to replicate it in part because it’s difficult to get the epidemiologic sample uh that would let you do that.  But in terms of this particular um phenomena, that study that we did um explicitly attempted to test this hypothesis of lag versus deficit uh on a large longitudinal uh epidemiological cohort and found quite clear evidence of a deficit as opposed to lag behind. 

Um, I actually went in and looked at each individual curve of the 400 so curves and we actually computed um fit statistics for each individual student and in um – in statistics one of the – the measures of fit that we use is sort of a percentage of variance accounted for by the model uh with the maximum being 100%.  And over 90% of the students, both normal and disabled students, the growth model accounted for upwards of 93% of the variance in their test score performance.  There was about 10% of the kids that the model did not do a very good job of capturing their growth data and so while we’re fitting a model that, you know, to you would look like – as time goes this way would look like this where kids are growing and then they’re sort of plateauing, uh and the – the non-disabled kids are growing and plateauing at a slightly higher level.  And the other thing that was interesting is that they actually all sort of plateaued around the same age, um, that there were some kids who sort of showed a growth curve that went like this and sort of had a little scalloping in it.  Um, but interesting enough they were pretty equally represented across the non-disabled and the disabled learners.  So I don’t know exactly what to make of it in terms of what might account for the fact that there’s a slightly different trajectory for those kids.  And again, it’s a relatively small percentage uh and didn’t appear to be related to their disability status.  Um, one of the things we don’t have in that study is really good evidence about the – any specific interventions that the kids might have been getting and we also at the time that that study was done early on, you know, these kids were identified in 3rd grade.  So we don’t have uh studies where – I’m sorry – in that particular study we don’t have information about really intensive interventions that might have been done with these students knowing what we know now.  Um, we do have other studies today.  For example, Joe Torgeson has done work where he has done really intensive phonologically based interventions with kids with reading disabilities uh early on and is able to actually change the trajectory that they’re on and then they stay on a uh – a more normalized trajectory and he’s been – he’s followed them for about 18 to 20 months, I think, and shown that once he gets them shifted up that they stay on a normal – a normal developing trajectory.  So it’s clearly the case that changes can be made with the right kind of interventions, but it’s not the case that those changes will just happen because students uh happen to mature and change naturally.

Actually – actually when I was referring to software I’m – I’m meaning it’s the wire – the wiring of the brain is the softwares.  How things are mod – you know, programmed and uh the intervention is the uh – is the process by which you reprogram.    

Well, phonetic, uh, the relationships among reading, and spelling, and phonetic segmentation, especially for young readers, I think are even stronger then for older readers uh because in older readers these processes become so fluent and automatic that uh you – you almost can’t uh, uh, see that they’re happening because students engage in them so – so automatically, unless they’re students with reading problems.  Um segmentation – phonetic segmentation is the process of taking a word and splitting it up into its sound parts.  Um and when you actually look at a word and you’re reading that word that’s basically what you’re doing is you’re splitting it up into its – its sound parts.  So this – this process of segmentation is very tightly linked to – to the ability to – to read um even though ultimately we get to the point where when we read words we perceive them as wholes.  In fact, we’re – that’s a trick of our – the automaticity of our perceptual phenomena.  We’re really processing these as serial connections of pho names but we’re so automatic and so fluent with it that it’s – it’s as if we’re processing it as holes.  And so one of the deceptive things is when you study advanced and skilled readers and you try to apply the models for skilled reading to developing readers and kids that are trying to acquire it, you’re – you’re mislead into thinking that, well, students should not worry about the sound parts of the words or the connections between the sound parts because basically – ultimately we want to get them as to where they see it as wholes.  I attribute that to being – like trying to take kids in Little League and make them Major League hitters without letting them go through the developmental process of learning how to hit off of a tee and then off of a machine.  And again, with spelling the connection is still the same in that it’s – it’s a different route to accessing the connections between the letters and the sounds.  It’s a different output production system.  So one of the things that we know from looking at uh a bulk of anecdotal evidence from around the country and some – and some experiments, is that when teachers teach writing and when they start teaching writing early, students read better.  So there’s a really strong connection between reading and writing, and in part, spelling is a small part of that.  I think as students become better readers they also become better spellers, especially if they’ve learned to read by being instructed in sort of the phonetics of – of our language they’ll be better spellers.  If they haven’t been given that kind of specific um phonetic instruction they’ll – their – they’ll be weaker spellers in that they will represent sounds phonetically but not necessarily the way we have chosen to represent them in our, you know, bizarre uh orthographic systems.  So – they’ll come up with a much more efficient coding system then the one we have.

Instruction is a really critical component to the prevention of reading problems and it gets at this notion that when we – when we do studies of students with reading problems – and there are a number of studies that are coming out now where we actually look at uh the organization of the brain and how the brain works in processing text, and we see that students with reading problems process text differently in terms of which parts of the brain are active and how they’re working together, but what – the other thing that we know is that when we provide the right kinds of interventions those connections in the brain change and they become much more like the way uh students without reading problems are processing the text.  So instruction is the process by which that change happens and when students get the right kind of instruction and they get the right kind of instruction early, it’s a lot easier to change those mappings then if you try to do this late or you allow problems to persist.  Because as you allow problems to persist students develop um their own systems and then it becomes harder to break those systems and introduce new systems uh then if you get started early and provide the right instruction early and students will develop those kinds of efficient uh decoding patterns quickly – uh early.  And students who struggle with phonological processes, if they get the right kind of instruction you – they can actually become fluent readers more easily if you provide that instruction early then if you wait until there’s been persistent failure and then try to, uh, change that process later.  You’ll – you’ll have to um work harder to make the changes and they will not become as automatic so the additional effort does not have the same degree or quality of payoff.  (Interruption)  Yeah, they can always improve.  You can always make a difference, but I’m – I’m trying to learn how to play the piano as an adult.  Well, there’s a limit to what kind of motor skills you can a – acquire as an adult and, you know, is – things are easier for kids, you know, especially motor – motor-based things and so…

Well, the preparation of – of – how we prepare young children to read well is – there are clearly some fundamental things we need to put in place for all kids.  Um not all kids require the same amounts of the same kinds of instruction so it’s really important to distinguish what students need at what stage of development and the fact that you could have 5 students sitting in the front row of your class and those 5 students are at very different stages of development with respect to reading.  They need different kinds of instruction.  And what got lost in the movement towards more implicit approaches to instruction was the fact that many students needed a much more systematic approach to linking up the connections between sound and letters and they didn’t make those connections on their own.  Or if they made them on their own it took longer for them to make them and they didn’t develop them to the same degree of skill and automaticity, and they would benefit from more explicit methods of instruction.  But what’s good about the movement, the implicit instruction method, not – not – it’s not the implicit instruction method that’s good, it’s that focus on enjoyment of reading.  It’s that focus on – reading is done for a purpose.  It’s done to extract meaning from print and it – it – it’s – it’s very important to keep that broader purpose of reading always in mind.  And for teachers who are teaching students about reading, it’s really critical that from the very beginning that they are focused on making meaning from what’s being read and that they’re building in their students the kind of language proficiency that will be necessary for students to be successful readers as texts become more difficult.  So as students move from learning to read to reading to learn, the kinds of language-based skills that students need to have are much greater and you can’t wait until 3rd and 4th grade to start trying to build those skills.  You need to be building those skills in kindergarten, 1st grade – pre-K, kindergarten, 1st grade, 2nd grade, 3rd grade.  So language development is a critical issue and that’s one of the really positive elements of the whole language movement.  But what got lost there was the focus on skills and the fact that the skills are an important part of this process as well.  So students need some grounding in this sound symbol relationship.  They need to automate that process while they’re – so that they can engage in the meaning process.  And so it was – you know, when the – when the balance is too far shifted in one direction or the other the process is not going to work.  So teachers need to understand that there are multiple components to reading – to – to good, effective reading instruction and uh there are documents out now that help teachers understand what those critical components are, how they come together, how you can identify where are students at in that process, uh and – so that you can tailor instruction so that you’re making sure that each kid – each child in your class is getting the kind of instruction that that child needs.

The issue about tailoring instruction and how we uh – where – where we’re at in terms of our diagnostic tools and our ability to map students to – to instructional methods given certain diagnostic tools, my sense is that scientifically and instructionally we know more and can do more then what we’re currently doing.  Um, one thing is having good instruction for all students in every classroom will eliminate, by prevention, a lot of reading problems that we see in the early grades.  Um, so simply by providing good, basic instruction we can prevent a large number of – of problems thereby sort of circumventing the need for more intensive assessment and – and sort of diagnostic-based interventions.  But we do know how to intervene with kids who have phonologically-based reading problems and there are sort of stair-stepped interventions where you can moderate them in terms of their intensity and their duration and monitor students’ progress towards the goals.  And we have at least 4 monolingual English-speaking children – we have good data on where kids need to be at kindergarten to have a good prediction of where they’re going to be at the end of 1st grade of 2nd grade.  We don’t have that same kind of information for second language children.  Even for uh second language children who are speaking a language that’s, you know, we’re – very prevalent then our culture Spanish.  We don’t have that same kind of information and we don’t know the extent to which – if we’re looking at their English language development and their reading in English if those same predictions hold up in the same way that they hold up for monolingual English speakers.  So the – when we talk about – when I’m answering this question I’m certainly focused on the – right now just on what we know about monolingual English speakers (unintelligible) the English language learners next, but I think we know more and can do more then what we’re currently doing in the schools and in part that’s because most of the interventions that have been demonstrated to be effective have been demonstrated to be effective in uh experimental settings and it’s only now that we know about them and we know how to help teachers learn how to deliver them or when is it the case that uh – that the intensity of the intervention is really something that needs to be done by someone outside of the classroom because of the – the amount of time involved.  And we do know a lot of things about just providing better classroom instruction, such as grouping students and uh – who are at similar levels of development.  And the teacher doesn’t need to have 15 groups in her classroom or his classroom.  You know, 3 groups is probably adequate in terms of being able to identify sort of clusters of students in terms of uh where they’re at in the reading development process, and then tailoring the instruction to those – to those small groups.  So I think – you know, I think the answer to the question is that while we don’t necessarily know everything that we need to know and there are certainly going to be some students who even given interventions that we know to be successful with most students are going to be not successful with those interventions.  In the research literature people refer to them as treatment resistors or slow to respond students.  And so we have – we’ve got work to do still in terms of knowing who those students are.  How can we identify them early?  If we had – if we could identify them early could we do something?  Could we provide a different sequence of instruction or maybe more intensive intervention earlier?  Would it be more effective?  So there’s a lot still that we don’t know, but not everything that we do know is being implemented in the – in the – in the schools and that’s partly uh where we’re trying to get to with, I think, a lot of this legislation and with the academies that are being done for schools and – and trying to get the people that are in the field working more closely with school districts and states to try to disseminate this knowledge and then helping to develop these – this kind of capacity within the – within the school systems.

Second language learners as it relates to literacy – my – the question was that did I have a soapbox that a – that I would – if I had a soapbox or what is my soapbox issue with respect to um, uh, second language learners and I probably have several.  Um, but I think the first is just our need for more and better longitudinal developmental research that looks at students who come to the country and are learning English as a second language um and what their – what the issues are that they’re dealing with and how do these relationships among things that we know about is being good predictors within monolingual English speakers?  How do those relationships change as a function of the context in which the students are growing up in?  How does that change in terms of uh the different kind of instructional environments that they’re growing up in?  How much does language proficiency in their primary language impact on these relationships?  What – to what extent does transfer occur from one language to another?  What’s the role of primary language instruction in learning to read in a second language?  We know there’s a role there but we don’t know precisely what that role is.  We need more and we need better longitudinal research to be able to answer these questions uh for these kids because they represent a very large percentage of our population and we don’t necessarily know what works bet or how to identify students early who are likely to have uh – have a difficulty and what the – what the predictions are.