Transcript for:
Understanding Research Methods in Human Development

howdy folks as we gear up to talk about the basics of research methods in human science i recognize that many of you will already know a lot about observational versus experimental research and so on but in this first lecture right now i'm going to tell you about some very specific ways human development science is a little different from what you normally see in other fields children are in an especially protected population in terms of research ethics and it's really important for you to understand the salience of why child development research methods are the way they are and what data from these specialized methods can and cannot tell us short story people with no training in human science and folks that's most people have a long history misunderstanding and therefore misinterpreting what outcomes from human development studies actually mean so we're going to clarify that you the tools you need to go out into the world and be part of the solution the many well-meaning but ultimately damaging policies the general public has about healthy child development you need to know this stuff because you're the ones who are going to go out and make a difference in the world let's start with this simple statement the best way to do research in child development conduct experiments true or false what do you think i want you to honestly reflect on this and see if you can identify where you got whatever belief you have about whether the statement is true or false was it science was it rumor was he read it think about the source what you think you know pause the lecture for a second and think about your answer okay most people would think this is true i mean experiments let us manipulate variables to identify cause and effect but that has to be the best way to do science right well in some sciences yes but in human science this is false and for a really good reason can yes start with an e ethics for example we'd like to know whether being abused as a child causes one to become an abusive parent later however to experimentally test this we would have to manipulate the variable of you got it being abused we have to randomly assign a bunch of kids to experimental or control groups and the kids in the experimental group would get abused and then we'd have to go back a couple of decades later and check on both groups and see how many of them became abusers and then compare rates between the two groups to see whether the kids we assigned to be abused became abusers more often than to control kids we assigned not to be abused wtf might we cannot do that instead we have to come at this question ethically and converge on the answer through multiple paths of observational and experimental research designs in many cases with child development variables the most accurate and ethical design is called the natural experiment and human development scientists use this one a lot so in a natural experiment it looks a lot like a true experiment because you have your comparison groups one of which does experience the variable you're interested in and one of which does not so like for the variable of abuse we can still track whether abused children are more likely to become abusers compared to kids who were not abused but the big difference is that we did not randomly assign those kids to be abused or not instead in a natural experiment we identify kids in the natural world who have already experienced our variable of interest or not experienced it and that's what determines whether they are in the natural experiment group or the control group so what's the problem with not having random assignment to groups the problem is it takes away our ability to control for compounds random assignment in a true experiment gives us reasonable confidence that all kinds of differences in people's characteristics that could affect our study outcomes those will be equally present in both experimental and control groups and so they'll effectively cancel each other out in a natural experiment groups are already created naturally by experiences we did not introduce so there could be something else that people in each group have in common that we don't know about so a confound is some variable or characteristic that's affecting our outcomes and it's not the one we're trying to study and we may not even know it's there or that it matters so for instance natural experiments do show that children who were abused are more likely to become abused appearance themselves compared with kids who weren't abused but those data also make it clear that only a relatively small percentage of abused kids actually become abusers wait what natural experiments show that only somewhere between 10 and 30 percent of kids who were abused grow up to become abusers the type of abuse and who the abuser was both seem to matter so in other words the majority of abused kids don't become abusers equally as important some adult abusers were not abused as kids these data show us that the experience of abuse does seem to cause an increase in a person's likelihood of becoming an abuser later but also that abuse alone cannot be the only cause when becoming an abuser otherwise all kids who were abused would become abusers and no unabused kids so taken together all these outcomes from natural experiments tell us there has to be at least one more variable in play that's driving a person to become an abuser and that unknown variable is what we call confound we have to keep making more natural experiments to look for connections between characteristics that look different in abusers versus non-abusers one such variable that looks like a promising candidate is a person's personality type particularly in terms of their degree of aggressiveness versus passiveness okay so let's look quickly at a kind of observational study that has a similar name okay to be crystal clear a natural experiment is considered a type of experiment whereas naturalistic observation is not considered an experiment at all it just describes what people are doing or animals or planets or chemicals or whatever what they're doing on their own naturally without any attempt to manipulate that behavior later scientists can look at all those descriptive data and see if there are any patterns that suggest group differences then they could design a natural experiment to test that but we don't start a natural observation study with any preconceived group differences in mind there's a whole other lecture called understanding observational studies where i talk about script methods like this one in more detail so i'm not going to repeat it here right now i'm bringing it up to make sure you're aware these two research methods are named similarly but they are quite quite different the last research related approach that specialized in human development science is not a methodology per se rather it's a set of different ways we select kids for any particular experimental or observational study one of the most common selection approaches is called cross sectional in this one kids from two or more different age groups are compared on some characteristic so maybe we want to know the average difference in the brain's white matter content between toddlers and teenagers so we measure white matter content in a group of three-year-olds and compare that to white matter content in a group of 16 year olds this one is used a lot let us collect data on age related changes all at once in a single study which makes this a fast and efficient approach however there are two important limitations to studies that use cross-sectional approaches because it's comparing groups of kids who are different ages first limitation is that we have to allow for the possibility that cohort effects could be confounding the data recall that cohort effects age group differences caused by the existing circumstances during the particular period of time that age group grew up and i talked about this in some detail on a previous lecture called what is a child the second limitation is that the cross-sectional approach can only tell us about average differences between age groups at that point like just a snapshot in time it can't tell us about the variety of individual differences there might be over time in whatever age-related change we're studying for instance in the white matter example the cross-sectional approach could only tell us that white matter content increases an average of say 200 percent from toddler to teenager it can't show us whether some individual kids increase 100 percent and others increase 300 percent into one for answers about the differences in ways kids can change on some age-related feature we need to test the same kids at a younger age and then test them again when they're older when we test the same kids at different ages we call this the longitudinal approach this approach eliminates the risk of cohort effect compounds so it can give us a cleaner picture about both the average change for the entire group over time as well as whether there are subgroups that change in a significantly different way than the average and this is important because then we can start studying why those subgroups are different than the average the longitudinal approach is definitely superior to cross-sectional in terms of quality of data and the breadth of outcomes but you see this one far less often why well practical reasons mostly to understand our white matter content question from earlier we would have to test a bunch of three-year-olds and then wait 13 years to test them again it takes a really long time to collect the data you need to answer your research plus you're going to lose some of your group people move away or die in a car accident or just don't want to come back and participate and study the second so longitudinal approaches tend to be expensive and time-consuming and they're not as common as cross-sectional ones okay so the last approach for selecting the kids we test is called microgenetic and it's sort of a compromise between cross-sectional and longitudinal approaches in microgenetic designs the same kids are tested repeatedly but only over a very short period of time usually a few days or weeks it tends to be used more with very young kids because all kinds of developmental changes are happening very quickly but testing them on some tasks say daily for a week gives us a lot of information about the change process for example microgenetic designs with three-year-olds measuring changes in the strategies they use to put together a new puzzle or learn how to read those designs show us the full picture of how the child strategy changes from before they know how to do something while they're learning it and after they've mastered it one of the coolest things i think microgenetic designs have uncovered a virtually universal response of just elation and delight preschoolers demonstrate when they have that aha moment where they figure out a new way to do things it just freaking blows their minds so cute and then of course they want to do the new thing five million times in a row well we'll see a lot more of this when we get to cognitive development okay medius that's it for this lecture don't forget to watch the rest of the memphis lectures and russia this material will be on exam one thanks and gig them