Brainerd, C. J. & Reyna, V. F., (2002). Recollection rejection: How children edit their false memories.

Brainerd, C. J. & Reyna, V. F., (2002). Recollection rejection: How children edit their false memories. Developmental Psychology, 38, 156-172.

Children’s memories can be easily falsified with both obvious and subtle suggestions from other people. Because of this, there is a concern in the validity of children’s testimony.

Even if an investigator is trained with the correct protocol, they repeatedly depart from the protocol and ask suggestive questions.
Children can be subjected to suggestions outside of the interview.
Endogenous distortion can falsify memory to some degree because these mechanisms are intrinsic to human memory.

**Children may possess cognitive editing operations that allow them to suppress the reporting of certain types of false memories (so interviewing procedures could be designed that encourage the use of editing operations).

Aim of the Present Article is to accomplish two things:

To show that recollection rejection is within the cognitive capabilities of young children.
To develop a new paradigm – "release from recollection rejection" that has two features:

Children respond to probes of memory with two types of instructional conditions (verbatim acceptance and gist acceptance)
A simple mathematical model to measure recollection rejection that uses performance in both of the above conditions.

Fuzzy Trace Theory

Verbatim and gist memories are deposited in parallel, so they are dissociated representations of experience. In memory tests, false-but-gist-consistent events are often erroneously recognized, but it may be possible for children to edit these events out of their memory reports if they can retrieve a verbatim memory. (A child remembers that he did not eat a hot dog, because he distinctly remembers eating a hamburger).

Verbatim traces are governed by accessibility and effectiveness of a cue. If young children are capable of recollection rejection, researchers should be able to reduce false recognition or false recall on gist-consistent events by manipulating accessibility of verbatim traces.

To test this, Brainerd & Reyna came up with the release from recollection rejection paradigm.

The memory tests were administered with verbatim acceptance or gist acceptance instructions. Verbatim acceptance instructions tell the child to select the item, only if it is the exact item they studied. Gist acceptance instructions tell the child to select the item if it maintains the meaning of what was presented, even if it doesn't exactly match what was said.
In the present studies children are tested with three types of items. For instance, imagine a child hears the sentence, "The coffee is hotter than the tea." A target is the exact item that was presented (e.g. "The coffee is hotter than the tea.") A true distractor is not the exact item that was presented, but maintains the meaning of what was presented (e.g. "The tea is cooler than the coffee."). A false distractor violates the meaning of what was presented (e.g. "The tea is hotter than the coffee.").
True distractors should be rejected under the VA instructions but accepted under the GA instructions. False distractors should be rejected under both instructions.

A simple mathematical model was developed that estimates recollection rejection in this paradigm. The model has four parameters.

V = The probability of retrieval of corresponding verbatim traces and the generating verbatim mismatches (using verbatim traces to make a decision). V indexes the process of recollection rejection.
G = the probability of retrieval of corresponding gist traces and generating meaning matches or mismatches (using gist memory to make a decision).
b_VA = acceptance using response bias for the verbatim instructions.
b_GA = acceptance using response bias for the gist instructions.

The model is described by a set of equations relating these parameters to observable responses of the subjects (i.e. the outcome space). The crucial equations are provided below: (1) P_VA,T(A) = (1-V)G + (1-V)(1-G)b_VA

** Kids will accept a true distractor under the verbatim instructions if they fail to retrieve a verbatim trace but do retrieve a gist trace or if they fail to retrieve a verbatim trace, fail to retrieve a gist trace, but accept the item due to response bias.

(2) P_VA,F(A) = (1-V)(1-G)b_VA

** Kids will accept a false distractor under the verbatim instructions only if they fail to retrieve a verbatim trace, fail to retrieve a gist trace, but accept the item due to response bias.

(3) P_GA,T(A) = V + (1-V)G + (1-V)(1-G)b_GA

** Kids will accept a true distractor under the gist instructions if they retrieve a verbatim trace or if they fail to retrieve a verbatim trace but do retrieve a gist trace or if they fail to retrieve a verbatim trace, fail to retrieve a gist trace, but accept the item due to response bias.

(4) P_GA,F(A) = (1-V)(1-G)b_GA

** Kids will accept a false distractor under the gist instructions only if they fail to retrieve a verbatim trace, fail to retrieve a gist trace, but accept the item due to response bias.

Brainerd and Reyna use these equations to algebraically solve for the model's parameters. Here are the equations for the model's parameters: V = [P_GA,T(A) - P_VA,T(A)] - [P_GA,F(A) - P_VA,F(A)]

Notice, the first part of this equation is analogous to process dissociation (i.e. R = I-E). The second part of the equation implements a model based correct for response bias.

G = [P_VA,T(A) - P_VA,F(A)] / (1-V)

Notice, that this too is analogous to process dissociation [i.e. F = E/(1-R)] with P_VA,F(A) correcting for response bias.

b_VA= P_VA,F(A) / [(1-V)(1-G)]

b_GA= P_GA,F(A) / [(1-V)(1-G)]

Measurement of Recollection Rejection in Children

Reyna & Kiernan (1994)

Using the memory test described above, R & K tested six- and nine-year-old children on eight narratives in a memory test immediately after the study period and one week after. They had three types of gist-consistent distractors that were each paired with gist violating distractors (TPN, FPN; TIO, FIO; TIN, FIN)

Using the mathematical model above, Brainerd and Reyna generated parameter estimates.

Immediate Test

Younger Children: V = .15 G = .24 b_VA= .37b_GA= .49

Older Children: V = .39 G = .31b_VA= .22b_GA = .52

The age increase in verbatim & gist processing was significant.

Pooled correct rejection rates for gist-consistent distractors

Younger: .57

Older: .71

On the delayed tests, age differences disappeared. The false alarm rate for gist-consistent distractors, however, remained stable in younger children, but raised from .29 to .45 in the older children.

Pooling across types of distractor and ages, the mean level of recollection rejection after one week decreased by 78% (from .27 to .06) and the mean level of gist processing decreased by 44% (from .30 to .17). Rejection became mostly passive nonacceptance because of a lack of memory support.

Model Validity Results – goodness of fit tests and parameter behaviors

To show that the paradigm was in fact measuring what Brainerd and Reyna said it was, goodness of fit tests were conducted (both qualitatively and quantitatively).

Qualitatively: Recollection rejection predicts that difference between acceptance rates for gist-consistent and gist-violating distractors must be greater in GA condition than in VA condition. This prediction was confirmed, (Figure 1, pg 162).

Quantitatively: Although the model they developed has exact algebraic solutions for each of the parameters, Brainerd and Reyna developed an interesting way to subject the model to goodness of fit tests. They created a new outcome space based on the joint probabilities of recognizing or not recognizing both an item's true distractor and false distractor. The equations for these joint probabilities were then subjected to goodness of fit tests using Hu’s General Processing Tree (GPT) program.

Parameter behaviors: Findings are consistent with fuzzy trace theory because the proportionate decline in V was nearly twice the proportionate decline in G over one week; AND the proportion of correct rejections of gist-consistent distractors (in VA condition) that were due to verbatim mismatches declined by more than 50%.

Kiernan, 1993

Kiernan used the same procedures as in the Reyna & Kiernan study, except: the developmental manipulation was different – instead of a difference in chronological age it was a difference in developmental delay, there were 16 narratives instead of 8, and only two gist-consistent distractors were administered (TPN & TIO).

Immediate Test

Control: V = .54 G = .39bVA = .44b_GA= .44

Delayed: V = .26 G = .27b_VA= .44b_GA= .48

_{Thus, control children were able to use verbatim
mismatch to suppress gist-consistent distractors about two times as much
as the delayed children.}

_{When pooling across developmental conditions,
there was a greater decline in recollection rejection (from .43 to .24)
than in gist processing (from .36 to .56, --an increase).}

_{Reyna & Kiernan (1995)}

_{The same basic study was conducted (with
children aged 6 and 9, with an immediate and delayed test), but instead
the narratives contained metaphorical statements about familiar objects
and events. There were three pairs of gist-consistent and gist-violating
distractors (paraphrases, perceptual meaning, and psychological meaning).}

_{Immediate Test}

Young: V = .08 G = .10 b_VA= .09b_GA= .40

Old: V = .45 G = .09b_VA= .03b_GA= .18

-Age increase in verbatim processing was significant in all three distractors. For the older children, there was a fivefold increase in the chance that correct rejection of a gist-consistent distractor was due to verbatim mismatch (.48 vs. .10)

Delayed Tests

Both age levels were no longer able to suppress false recognition of gist-consistent distractors with verbatim mismatch. (It was not significantly greater than zero for any of the six conditions except TPSY in 6 year olds = .07). There was a little decline in gist processing for 6 year olds, (.10 to .06) and the mean value of gist processing for nine year olds nearly doubled.

False alarm rate for Immediate to Delayed Test

Younger : .17 to .34 (it doubled) Older: .07 to .49 (seven-fold increase)

As in the other Reyna & Kiernan study, after one week, rejection of gist-consistent distractors became largely a factor of passive nonacceptance due to lack of memory support.

Model Validity Results

Goodness of fit for qualitative data, Figure 3, pg. 166.

The evidence for quantitative data was the same as in the Reyna & Kiernan (1994) study.

Both qualitative and quantitative data supported the paradigm.

General Discussion

Five main instructive findings:

- Dramatic developmental improvements in recollection rejection (both in chronological development and mental development).

- Despite developmental differences, evidence was given that suppression of false acceptance of gist-consistent events is possible at the earliest developmental level tested (which means that even younger children might be able to do this).

- Both qualitative and quantitative fit tests showed that the recollection rejection model gave a good account of the three data sets (confirming that the release-from-recollection-rejection paradigm is an appropriate tool for studying false memory editing in children).

- Parameters V & G reacted to manipulations in ways that are congruent with theoretical definitions.

-Parameter V displayed more developmental variability than Parameter G.

Use in Forensic Interviews:

Two possibilities were found to exploit the recollection rejection operations and the release from recollection rejection paradigm to improve forensic interviews.

-Use in the prediction of new procedures to reduce false memory reports.

-Use release-from-recollection-rejection paradigm to

Pinpoint actual memory effect of forensic interviewing

Measure actual memorial bases for individual differences in false reporting (by estimating V, G, and b for individual children)


University of Arkansas	Department of Psychology	Graduate Program in Experimental Psychology
Lampinen Lab	False Memory Reading Group	False Memory Reading Group Fall 2002