Hannigan & Reinitz propose that conjunction errors occur due to problems of binding information at study and / or retrieval. For example, if at study participants see the words Blackbird and Jailhouse, the configural information used to bind the word Black & bird together and Jail & house together may dissolve. Thus, at retrieval the parts of the word Jail & bird are recombined to form the new conjunction Jailbird. In contrast, Jones & Jacoby (2001) claim that familiarity-based retrieval accounts for memory conjunction errors. For example, the conjunction Jailbird invokes a feeling of familiarity because both parts of the word were studied, though not in the particular combination as presented at test.

The following experiments examine whether source monitoring for context of presented stimuli can reduce memory conjunction errors. Specifically Marsh et. al examined:

1. Will a source monitoring test reduce conjunction errors?

2. Must the parent stimuli from which conjunction stimuli are formed be studied

from different sources for a reduction of conjunction errors to occur.

3. What combinations of sources can reduce conjunction errors. Specifically, will

Methods: Participants included 80 undergraduates. 99 compound word triplets were created and denoted as A-C. A & B words were the parent stimuli (deadbolt & neckline) and C was the tested conjunction formed from the parent words (deadline). Half of the words were seen and half of the words were presented auditorally.

Same source conditions: A & B compound words were presented in the same modality.

Different source conditions: A & B compound words were studied in different modalities. At test participants in the recognition condition were asked to indicate if the word was old or new. In the source monitoring conditions, participants were asked to indicate if the word was seen, heard, or new.

Results: Source monitoring did not reduce conjunction errors & conjunctions formed from different contexts at study were not avoided any better than when presented in the same context. Participants were more accurate identifying the source of conjunctions when the parent stimuli were presented in the same modality. Conjunctions were indicated to be heard more than seen.

Experiment 2 examines internal-internal source monitoring on conjunction errors. Method: The same triplets were used as Experiment 1. At study however, a total of 80 participants were asked to indicate on a scale of 1=7 How Imageable OR How Categorizable the word was. At test, participants in the source monitoring task were asked to indicate if the word was new, imagined, or categorized.

Results: Source monitoring did not reduce conjunction errors nor did the same or different context in which conjunction words were formed affect conjunction errors. Participants had a higher hit rate for old words when they used cognitive process of imagining or categorizing the study word. Participants identified the source of conjunction words more accurately when the conjunctions were made of parent words studied in the same modality.

Experiment 3 examines monitoring with internal-external combination of sources.

Method: 80 participants. The procedures were the same as the previous two experiments with the following exceptions. At study, participants were visually presented with anagrams which participants solved, or the parent word was presented auditorally. Source monitoring at test was the same as the previous experiments & conjunctions were formed from words that were presented in the same and different modalities.

Results: Source monitoring didn’t reduce conjunction errors and in fact increased conjunction errors when parent words were studied in the same modality. Conjunctions formed from the parents that were presented in the same modality were more often correctly indicated as to the correct context in which the parent words were studied. As in Experiment 1, participants labeled more conjunctions as having been heard than generated.

Experiment 4

Method: 80 Participants followed the same procedure as in Experiment 3 with the following exceptions. Experiment 4 replaces the heard source with the seen source used in Experiment 1 to make source monitoring more difficult. At test, participants in the source monitoring condition were asked to indicate if the word was seen, generated, or new.

Results: Participants made more conjunction errors in the source monitoring condition than the recognition condition. Conjunctions from the same source were claimed to be old more than conjunctions from different sources, but this was not a significant difference. More conjunctions were labeled as seen than generated.

General Discussion:

Marsh et. al indicate the results of these experiments support Jones and Jacoby’s (2001) theory that conjunctions are called old because of the familiarity conjunctions provoke, however another process must also be going on due to the difference of conjunction errors in Experiment 3 & 4.