“I remember thinking …”: Neural activity associated with subsequent memory for stimulus-evoked internal mentations

Abstract

Conscious thought involves an interpretive inner monologue pertaining to our waking experiences. Previous studies focused on the mechanisms that allow us to remember externally presented stimuli, but the neurobiological basis of the ability to remember one’s internal mentations remains unknown. In order to investigate this question, we presented participants with sentences and scanned their neural activity using functional magnetic resonance imaging (fMRI) as they incidentally produced spontaneous internal mentations. After the scan, we presented the sentences again and asked participants to describe the specific thoughts they had during the initial presentation of each sentence. We categorized experimental trials for each participant according to whether they resulted in subsequently reported internal mentations or not. The results show that activation within classic language processing areas was associated with participants’ ability to recollect their thoughts. Activation within mostly right lateralized and medial “default-mode network” regions was associated with not reporting such thoughts.

Keywords:

• Self
• Subsequent memory
• fMRI
• Mind wandering
• Consciousness
• Language

Notes

¹  Despite the relatively very few remaining participants (13), the pattern of activation was markedly consistent across all participants, leading to the observed statistical robustness.

²  We did not have any a priori prediction regarding insular activation, and therefore this finding will not be further discussed. It appears in Table 1 for future reference.

³  In Study 1, we omitted participants who were unable to report 25 mentations out a total of 312 trials (i.e., 8% of the trials). In the current experiment, the total number of non-lure trials was 180 and so maintaining an 8% proportion entailed a cutoff of 15 mentations. The resulting proportion of remaining participants (65%) was identical in both studies.

⁴  While we employed a standard threshold of p < .001 with 60 contiguous voxels (corresponding to a corrected p-value of .05), a large portion of the default-mode network also survived a much more stringent threshold of p < .00005, k = 60.

Additional information

Notes on contributors

Michael Gilead

Michael Gilead, Nira Liberman, and Anat Maril designed research and wrote the article; Michael Gilead performed research; Michael Gilead and Anat Maril analyzed data.

Nira Liberman

Michael Gilead, Nira Liberman, and Anat Maril designed research and wrote the article; Michael Gilead performed research; Michael Gilead and Anat Maril analyzed data.

Anat Maril

Michael Gilead, Nira Liberman, and Anat Maril designed research and wrote the article; Michael Gilead performed research; Michael Gilead and Anat Maril analyzed data.