In this unit we have considered many experimental studies in which two or more groups or conditions are compared. Some of these experiments are conducted under strictly controlled conditions which increases internal validity but reduces the effect of confounding variables on the dependent variable. This type of study is usually referred to as a laboratory experiment, although the highly controlled setting does not always need to be an actual laboratory, it could be a room in a school or community centre so long as the researchers have strict control over situational variables such as the placement of furniture, lighting, sound and any other stimuli that might affect performance. 

Despite being highly scientific, such studies lack ecological validity. Children and their parents are less likely to behave naturally when under such circumstances, especially when they know they are being observed. This is why researchers sometimes prefer to conduct field experiments, where an independent variable is manipulated but the data is gathered in a natural setting, such as the child’s home or school. Examples of different types of experiment are provided in table X. 

Activity: For each experiment in the table:

• identify the independent and dependent variables;
• write a directional and a non-directional experimental hypothesis and a null hypothesis;
• identify the experimental design (independent groups, repeated measures or matched pairs);
• Can you think of any situational factors that might have been controlled in Goldstein et al. and Song et al.? Can you think of any situational factors that may not have been controlled in Schonert-Reichl and Lawlor (2010)? 

Table 1:

Setting	Study
Laboratory	Goldstein et al. (1990). See page XX.	Investigated shaping in infant language development. The researchers compared the complexity and maturity of infant babbling in two groups of eight month old babies. In the contingent group, the mothers smiled and moved towards the babies every time the baby made a sound. In the non-contingent group the mothers smiled and moved towards their babies but these reactions were not matched to their babies’ behaviour.
Laboratory	Song et al. (2010). See page XX.	Investigated the role of different features of infant directed speech (parentese) in 19 month old babies. Babies were asked to look at pictures. On half of the trials they were asked to look at the target picture using typical parenteses. On the other half of the trials, the instructions were given by a digitally altered voice in which properties of parentese such as speaking rate and pitch were altered.
Field	Schonert-Reichl and Lawlor (2010). See page XX	Investigated the effectiveness of Mindfulness Education (ME) programmes in Canada. The ME group was compared with a matched wait-list control group who did not participate in ME. Teachers rated all the children for socially competent classroom behaviour.

Each of the studies in Table 1 uses a different experiment design, all of which have associated strengths and weaknesses. For example, unless there is random allocation to groups, the internal validity of findings from independent groups designs may be reduced by participant variables. If Goldstein et al. (1990) failed to randomly allocate their babies, it would be impossible to know whether differences in the maturity of their babbling was due to the manipulation of the mothers’ behaviour or individual differences in the rate of language development. This is why researchers often test participants in independent measures designs studies before the experimental manipulation to demonstrate that the two groups are similar and therefore comparable.

In the study by Song et al. the researchers randomised the order of the trials for each baby. This meant that different pictures were paired with the normal ‘parentese’ voice or the digitally altered voice for each baby, i.e. for each target image there was an equal chance that it would be presented with the normal or digitally altered voice. This is an important way of overcoming order effects. If the first half of the images were always presented with the normal voice and the second half with the altered voice it would be impossible to conclude that the babies performed more poorly with the altered voice due to the lack of certain important features of parents. It may simply have been due to the fact that by the second set of pictures the babies had become bored or tired. Another way that order effects can be controlled is through the use of counterbalancing.

Activity: How could you change the experimental design of each study to improve internal validity? For example, how could you turn Goldstein et al (1990) into a matched pairs design? How could you use counterbalancing instead of randomisation in Song et al. (2010) How could you change Schonert-Reichl and Lawlor (2010) so that they used an active control group rather than a wait-list control group?