Hirshman, E. (2004). Ordinal process dissociation and the measurement of automatic and controlled processes. Psychological Review, 111, 553-560.

 

What is process dissociation theory?

 

Process dissociation theory is a model of how conscious and unconscious memories are related as well as a measurement technique designed to tease apart conscious and unconscious influences.  In the process dissociation technique participants answer recognition memory test items under either inclusion or exclusion instructions.  Inclusion instructions ask participants to accept items regardless of their source, exclusion instructions ask participants to accept items from one source but reject items from another source.

 

Example, say you watch a video of an office theft. You then read a narrative describing the theft that includes a number of misleading suggestions (e.g. there was a calculator on the desk when in fact there wasn't).  Inclusion instructions would require you to select both the items that were in the video and the items that were in the narrative (so you should say 'yes' to the calculator).  Exclusion instructions would require you to say yes only to those items from the video (so you should say 'no' to the calculator).  According to the process dissociation approach, the probability of accepting the critical items (e.g. calculator) in the two instructional conditions are given by the following equations:

 

I = C +(1-C)A

E = (1-C)A

 

Where I is the probability of accepting an item under the inclusion instructions, E is the probability of accepting an item under the exclusion instructions, C is the probability of a controlled influence of memory (i.e. recollection), and A is the probability of an automatic influence of memory (i.e. familiarity).

 

 

It's a matter of simple algebra from these equations to get estimates of controlled and automatic influences of memory:

 

                C = I - E

                A = E/(1-C)

 

So if you are willing to accept the model, process dissociation provides a nice way of estimating the relative contributions of controlled and automatic processes to memory.

 

Critiques of Process Dissociation

 

A number of people have provided critiques of the process dissociation approach and two of the more important critiques have been the following:

 

Independence: Process dissociation assumes that the probability of controlled and automatic processes are independent, but researchers have argued that is not the case.

 

Invariance: Process dissociation assumes that the probability of controlled and automatic processes are not influenced by what instructions people are given, but researchers have argued that is not true.

 

The process dissociation equations only work if both independence and invariance are correct assumptions.

 

Hirshman's Basic Assumptions

 

Basically what Hirshman is doing is trying to develop a way of drawing conclusions from inclusion and exclusion results without having to make the strong assumptions that the process dissociation procedure currently makes.  The equations for inclusion and exclusion developed by Jacoby wouldn’t hold anymore, but maybe you could still draw some interesting conclusions from exclusion and inclusion conditions. So, he starts with a fairly simple set of assumptions:

 

  • For any value of A, if you increase C you will increase I
  • For any value of C, if you increase A you will increase I
  • For any value of A, if you increase C you will increase E
  • For any value of C, if you increase A you will decrease E
  • Lastly, make the assumption that both A and C are greater than zero and less than one      

 

Possible Experimental Results

 

What Hirshman wants to do is be able to take results of inclusion and exclusion conditions and draw conclusions about what they mean while making the fewest possible theoretical assumptions.  He concludes that even without making all of Jacoby’s assumptions you can still draw a number of useful ordinal level conclusions from process dissociation experiments. Hirshman says, imagine you are conducing an experiment and there are two conditions.  You get inclusion and exclusion results from each condition. What are the possible ways the experiment could turn out, and what would those possible results tell you? So here are the possibilities he considers:

 

Data Pattern I

 

E1 = E2

I1 > I2

 

Data Pattern II

 

E1 > E2

I1 = I2

 

Data Pattern III

 

E1 > E2

I1 > I2

 

Data Pattern IV

 

E1 > E2

I1 < I2

 

Given that these patterns are the ones of greatest interest, Hirshman next sets out to decide what you can conclude if you got one of those four patterns in your data.

 

Conclusions Given the Various Data Patterns

 

Data Pattern 1.

 

If E1 = E2 and, I1 > I2, then  A1 > A2 and C1>C2

 

If  I1 > I2 then there are three possibilities:

 

(1)      (1)      A1 > A2

(2)      (2)      C1>C2

(3)      (3)      Or A1 > A2 and C1>C2

 

These possibilities are further constrained though when you are told E1 = E2 .  If it were only the case the A1 > A2 then E1 > E2 .  If it were only the case that C1>C2 then E1 < E2 .  The only way you could get E1 = E2 is if the automatic and controlled processes offset each other.

 

Data Pattern 2.

 

                If E1 > E2 and, I1 = I2, then  A1 > A2 and C2>C1

 

If I1 = I2, then  there are two possibilities:

 

(1)      (1)      A1 = A2 and C1=C2

(2)      (2)      Or A1 > A2 and C1<C2

 

However, we know that E1 > E2 so it can’t be the case that there has been no change in either A or C. Rather, they must have changed in offsetting ways.

               

Data Pattern 3.

 

If E1 > E2 and, I1 > I2, then  A1 > A2 and results for controlled processes are indeterminate

 

In this case we have a independent variable that is increasing responses in both the inclusion and exclusion conditions. Consider first the exclusion instructions.  If E1 > E2 then either A1 > A2 or C1 < C2 or both.  However, if C1 < C2 and that were the only thing going on, then I1 < I2, so it must be the case that A1 > A2

 

Data Pattern 4.

 

                If E1 > E2 and, I1 < I2, then  C1 < C2

 

Essentially one has a crossover interaction here where the IV has one effect on the exclusion condition and the opposite effect on the inclusion condition.  Because automatic processes push both I and E in the same direction, this pattern cannot be produce by changes in A alone. Rather, it must be the case that the controlled processes are being influenced by the independent variable pushing the I condition in one direction and the E condition in the opposite direction.

 

Discussion

 

Hirshman compares his approach to the original process dissociation approach. Like, process dissociation, Hirshman points out that his approach makes use a a logic of opposition. However, the approach does not assume that recollection and familiarity are independent. The approach makes only the bare essentials in terms of assumptions about the relationship between controlled and automatic processes.

 

 

University of Arkansas

Department of Psychology

Graduate Program in Experimental Psychology

Lampinen Lab

False Memory Reading Group

False Memory Reading Group Fall 2004