Lecture 7

Question 1

Describe an AB design. What are potential issues with these designs?

Answer 1

A is the baseline phase, and B is the treatment phase. A acts as a control condition to show a treatment effect.

Internal validity is not strong, as the measures obtained in the control phase aren’t concurrent with measures in the treatment phase. Therefore, some other factor may have caused the change.

Question 2

Describe the ABA design. What’s an advantage of the ABA design over the AB design?

Answer 2

ABA designs measure a baseline phase before AND after treatment. Internal validity is increased, by demonstrating that improvement coincides with treatment.

HOWEVER this is not particularly useful when considering that we hope clients would experience a lasting change

Question 3

What is the ABAB design?

Answer 3

This single case design shows that intervention works with the same behaviour a second time, thereby increasing internal validity. However, this design assumes that treatment effects are reversible (and therefore isn’t applicable to most SP intervention)

Question 4

What is a multiple baseline design?

Answer 4

Treat one client/behaviour at a time.
E.g. the clinician chooses 3 behaviours, and introduces treatment for the behaviours in a staggered fashion. While one treatment is introduced, the client is measured on the control conditions (i.e. other areas of language) CONCURRENTLY - increases internal validity

Question 5

What is the ABCB design?

Answer 5

A = baseline
B = treatment 1
C = treatment 2

It is difficult with these designs to tease out the carry-over effects (i.e. is the client only showing a change in C because B was presented first?)

Question 6

What is a crossover design?

Answer 6

Two treatments, X and Y, are applied to different modalities of language, assuming that both deficits have the SAME underlying aetiology.
O1 - X - O2 - O3
O1 - O2 - Y - O3
While one area of language is being treated, another acts as a control. Then the treatments are “crossed over”

Question 7

What are advantages of a crossover design?

Answer 7

Addresses ethical dilemma of withholding treatment
Addresses spontaneous recovery - if it were true, we would expect to see change in O1 to O2 for BOTH conditions. Therefore if Y does NOT improve from O1 to O2, then we can argue that spontaneous recovery has not occurred.

Question 8

What are three ways of introducing a control condition in therapy?

Answer 8

1. Comparing a set of treated ITEMS with a set of untreated items.
2. Comparing a treated MODALITY (e.g. phonology) with an untreated modality (e.g. syntax) that have the same underlying aetiology.
3. Use the same behaviour as its own control (as in AB or ABA designs).

Question 9

What is an issue with comparing a set of treated items with untreated items?

Answer 9

While it is actually DESIRED that a client generalise a skill/behaviour to new items, we must show that there is NO improvement. If there is improvement, then it can be argued that the presence of another variable caused this change.

Question 10

What is an issue with comparing a treated modality with an untreated modality?

Answer 10

It is difficult to ensure that the SEVERITY of each modality is EQUAL, and that tasks are of a similar difficulty.

Question 11

What is an issue with using the same behaviour as its own control?

Answer 11

Measurements are not concurrent, and therefore internal validity is low. Something else may have coincided with the treatment to induce change.

Question 12

What are two methods of visual analyses of time-series data?

Answer 12

1. Plotting the celeration line and conducting a binomial test to reveal if treatment has significantly increased the level of responding
2. 2-SD band method. The mean, -2SD and +2SD are plotted for the baseline phase (provided it has NO SLOPE). These lines are extended to the treatment phase, and if values from the treatment phase extend beyond 2SD’s above the baseline mean, then a treatment effect is supported

Question 13

What does the Mann test do?

Answer 13

Evaluates whether there is an increasing or decreasing relationship between the DV and the session number.
p = number of values to the right of a session that are LARGER
q = number of values to the right of a session that are SMALLER
P = sum of p’s
Q = sum of q’s

S = P - Q (as S gets further from 0, we can assume that treatment is having a real effect, as subsequent sessions are either MOSTLY larger or MOSTLY smaller - a TREND (not random)