This is the third in a series of articles we've read by Todd Jones on the topic of memory conjunction errors (Jones & Jacoby, 2001; Jones, Jacoby & Gellis, 2001).  Memory conjunction errors happen when people falsely remember items (e.g. heartache) that are made up of a mixture of features from presented items (e.g. heartburn + toothache).  The two presented items (e.g. heartburn and toothache) are called parents and the lure made up of those features (e.g. heartache) is called a conjunction lure.

The authors describe four competing theories for how memory conjunction errors happen:

• Single process familiarity: Both true memories and false memories occur based on familiarity, where familiarity is some function of the the number of features in the test item that match items in memory.
• Dual process account: Recognition memory is a joint product of recollection and familiarity.  Both recollection and familiarity can produce true memories.  However, recollection tends to reduce false memories working in opposition to familiarity (see Brainerd, Reyna, & Mojardin, 1999; Brainerd & Reyna, 2002; Clark & Gronlund, 1996; Hintzman, Curran & Oppy, 1992; Rotello & Heit, 1999; 2000; Rotello, Macmillan & Van-Tassel, 2000).
• Configural account (Reconstruction): Recognition memory is a joint product of featural information and configural information.  Memory conjunction errors reflect a reconstructive process in which features are errantly bound together at retrieval.
• Configural account (Construction): Like the reconstructive view, except that the feature binding errors are assumed to occur at study rather than at test (see Hannigan & Reinitz, 2000; Reinitz & Hannigan, 2001).

Using this paradigm they manipulated the lag between seeing the parents (e.g. heartburn, toothache) and seeings the conjunction (e.g. heartache).  The idea is that if there are only a couple words separating the parents and the conjunction lure, participants should be more likely to consciously recollect the parents and reject the conjunctions (e.g. I know it wasn't heartache, it was heartburn).

Methods

Students saw 256 compound words  in a continuous recognition test.

• Targets: Words that had been previously seen
• Conjunctions: Words that were made up of features from previously seen words
• New: The first time a word is presented (as long as its not a conjunction)
• Fillers: Irrelevant words that were repeated

The lag between the parents and the conjunctions was varied.  In Experiment 1 the lags were 1, 5, & 20.  In Experiment 2 the lags were 0, 1, and 5.  Keep in mind what this means.  Zero lag means the words are presented back to back (e.g. Heartburn is presented then immediately afterwards Heartache is presented).

Participants were given specific instructions about how to respond:

• They were told if they were unsure to always guess "old"
• They were told what conjunction lures were and that they would be on the test
• They were told specifically that if the test item they were evaluating included a part that had occurred in a previous word, they could be sure that the test item was new.  (e.g. If you saw heartburn then heartache must be new).

Experiment 1: Conjunction Errors

Lag of 1	Lag of 5	Lag of 20
.51	.46	.39

Experiment 2: Conjunction Errors

Lag of 0	Lag of 1	Lag of 5
.37	.46	.44

Discussion

Here are a few of the points the authors made in the general discussion:

• The dual process account didn't get very strong support.  In order to produce any evidence for recollection rejection at all they had to actually present the items back to back.
• They raised the possibility that maybe this result is unique to the continuous recognition paradigm.  Maybe if every item is a test, you don't really encode it very well.  This, they said, would be like divided attention which we know limits recollection.
• They talked about how one could really combine the dual process and configural account into a single account.
• They also talked about whether it really makes sense to talk about configural information when the stimuli one is using are polysyllabic or polymorphemic words.