Distinct deficits of repetition priming following lateral versus anteromedial frontal cortex damage

Abstract

Object repetition commonly leads to long-lasting improvements in identification speed and accuracy, a behavioral facilitation referred to as “repetition priming”. Neuroimaging and non-invasive electromagnetic stimulation studies have most often implicated the involvement of left lateral frontal cortex in repetition priming, although convergent evidence from neuropsychological studies is lacking. In the current study, we examine the impact of surgical resection for the treatment of epilepsy on the magnitude of repetition priming at relatively short-term (30–60 minute delay) and long-term (3 months) delays in 41 patients with varying seizure foci and resection locations. Overall, patients exhibited significant repetition priming at both short-term and long-term delays. However, patients with frontal resections (largely anterior and medial frontal) differed significantly from those with right anterior temporal resections in showing fully intact short-term priming but absent long-term priming. In a comparison set of 10 recovered aphasic patients, patients with left lateral frontal damage exhibited impaired short-term priming relative to other frontal damage locations, suggesting the differential involvement of lateral and anteromedial frontal regions in mediating repetition priming at short-lag and long-lag timescales, respectively.

1. Introduction

Repetition priming is an observed behavior in which repeated exposure to a stimulus leads to an increase in identification speed and accuracy. This behavioral phenomenon is a major form of implicit memory, as it occurs automatically (e.g. Logan, 1990), it is obligatory (e.g. Wiggs et al., 1997), and does not require conscious awareness (e.g. Cave and Squire, 1992). It is also extremely long-lasting, having been observed over weeks, months, and even years (e.g. Cave, 1997; Mitchell, 2006; van Turennout et al., 2000; Wiggs et al., 2006). Repetition priming is not dependent on the hippocampus (e.g. Cave and Squire, 1992), suggesting instead a dependence on the neocortex (e.g. Demb et al., 1995; Gotts, 2016; Maccotta and Buckner, 2004). In terms of neural mechanisms, repetition priming has been associated with “repetition suppression”, a decrease in the neural response to a stimulus after repeated exposure (e.g. Schacter and Buckner, 1998; Wiggs and Martin, 1998), as well as increased functional and effective connectivity between brain regions (e.g. Ghuman et al., 2008; Gotts et al., 2021). The mechanistic link between repetition priming and repetition suppression has not been determined, nor has the complete functional neuroanatomy of the two phenomena (e.g. Gotts et al., 2021, 2012; Grill-Spector et al., 2006). The regions of cortex associated with repetition priming and repetition suppression will vary based on the task and the processing requirements, including stimulus modality and task demands (e.g. Dobbins et al., 2004; Maccotta and Buckner, 2004).

One area of neocortex for which neural activity has been implicated in repetition priming is the left lateral frontal cortex, particularly in the territory of the inferior frontal sulcus. Studies using both neuroimaging (e.g. Dobbins et al., 2004; Ghuman et al., 2008; Gotts et al., 2021; Horner and Henson, 2008; Maccotta and Buckner, 2004) and transcranial magnetic stimulation (e.g. Wig et al., 2005) have suggested not only the correlational but causal involvement of left frontal cortex. However, given the potential for poly-synaptic spread in transcranial magnetic stimulation (e.g. Bergmann and Hartwigsen, 2021; Cowey, 2005; Logothetis et al., 2010), convergent evidence from other methods such as neuropsychology is also important in assessing causality (see also Martin and Gotts, 2005, for discussion). There have been very few studies in the neuropsychological literature that demonstrate a relationship between repetition priming and focal brain damage. A study by Swick (1998) examined 11 patients with lesions centered in the left dorsolateral prefrontal cortex. The patients were divided into two groups depending on whether the lesion was more anterior or posterior in frontal cortex. A lexical decision task with visual stimuli repeated after a delay showed no deficit in behavior of repetition priming in the patient group compared to controls, only an attenuated event-related potential for repetition effects. A further study by Gabrieli et al. (1995) described a patient (MS) with impaired repetition priming but intact recognition memory for visually presented words following resection of the right occipital cortex for the treatment of epilepsy, with the reverse pattern (intact repetition priming but impaired recognition memory) observed for two amnesic patients (due to Korsakoff’s syndrome in one and temporal lobe epilepsy in the other). Three additional epilepsy patients with resections in left occipital, right frontal, and left frontal cortex and four stroke patients with lesions in either right temporal or left parietal cortex exhibited intact repetition priming for visually presented words when considered as a group. More neuropsychological studies are needed to further assess how repetition priming might be mediated by frontal regions, as well as regions outside of frontal cortex.

The current study addresses some of these needs by examining patients with epilepsy who are candidates for resection of cortex before and after the surgical intervention. Forty-one patients with a range of assessed seizure foci performed an object naming task prior to and three months after a direct resection of cortex, with a subset (five) of these patients having resections in frontal cortex and other subsets having resections in the left and right anterior temporal lobes, 17 and 10 patients, respectively (9 patients had more idiosyncratic removals, an “Other” group, that could not be easily grouped). In order to examine repetition priming, patients were presented with a group of objects to name, and then after an approximately 30–60 minute delay named previously encountered objects along with a group of novel objects. Three months after cortex resection, the patients again named objects presented after a ~30–60 minute delay, as well as objects they had named in the session prior to resection (“3-month delay”) along with a set of novel objects. Testing with the same paradigm before and after resection allows each patient to act as their own control with regard to the role the removed section of cortex plays in behavioral repetition priming. Due to the relatively few patients with resections specific to the left lateral frontal cortex (that has been most associated with repetition priming), we also examine repetition priming in 10 recovered aphasic patients with extensive left-hemisphere frontal cortex damage due to stroke. These patients participated in one object naming session with a ~30–60 minute delay between presentations of novel and repeated stimuli. Patients in the “Other” group are included for analyses that test performance across all patients and prior to surgery, but they are excluded for the post-resection analyses of the effect of group on priming (since they do not represent any consistent location of damage for which the results would be interpretable).

2. Materials and methods

2.1. Participants

Forty-one patients (17 female) with drug-resistant focal epilepsy who were candidates for cortical resection participated in the object naming study. The mean (SD) age was 35.2 (11.8) years (range: 19 to 58). Patients were tested for handedness (37 right, four left), language dominance (39 left, one right, one bilateral), and full-scale IQ (mean=85.6, SD=14.0, range: 53 to 120, N=38). The average age for the onset of seizures was 19.0 (14.0) years (range: 0 to 50). All patients received a full neuropsychological evaluation from a trained clinician prior to resection. Tests administered and scored during the evaluation were examiner and circumstance dependent. The tests and select sub-tests with at least 73% of patients scored (N=30) includes the WASI (Wechsler, 2011), EIWA (Wechsler, 2008a), or WAIS (Wechsler, 2008b) Full Scale IQ, Verbal Comprehension Index, Perceptual Reasoning Index and Digit Span, the WMS (Wechsler, 2009) Logical Memory Immediate and Delayed Recall, the BNE Verbal Memory – Prose Immediate and Delayed Recall (Artiola i Fortuny et al., 1999), the CVLT (Delis et al., 2000) Total Correct and Short Delay Free and Cued Recall, the NAB (White and Stern, 2003) Language Module Auditory Comprehension Test and Naming Test, and the Boston Naming Test (Kaplan et al., 1983). Four patients were tested with the BNE Verbal Memory – Prose subtest in place of the WMS Logical Memory subtest. Seven patients were scored on both the NAB Language Module Naming Test and Boston Naming Test, while 22 patients received the NAB Naming Test only and nine received the Boston Naming Test only. Since both tests are scored as a normalized T-score, naming test scores were combined, with NAB Naming score used if both were available. Thirty-four of the patients were given a full neuropsychological evaluation 1-year post resection. In most cases, the same tests were administered as in the pre-operative evaluation. When giving responses during the study task, seven participants were non-English speakers and gave their verbal responses in their native language (six Spanish, one French). Seven other participants were non-native English speakers but gave their responses in English. The other 28 participants were native English speakers.

Ten patients (five female) that had frontal lesions due to MCA stroke participated in a separate experiment using only the relatively short-term delay (~30–60 minute delay). The mean (SD) age was 58.5 (10.4) years (range: 36 to 69). The mean number of years since the participant’s stroke was 10.8 (3.4) (range: 6.6 to 16.2). These participants were given a neuropsychological examination that included the WAB (Kertesz, 1982) to assess residual language abilities, the TEA (Robertson et al., 1994) to assess executive functioning and attentional abilities, and the CVLT and “Doors and People” assessment (Baddeley et al., 1994) to assess verbal and visual memory abilities.

All subjects gave informed consent and were compensated for their participation. Ethics approval for this study was granted by the NIH Institutional Review Board (clinical trials numbers NCT01273129 and NCT00001308).

2.2. Stimuli and Task Design

Stimuli for object naming consisted of 200 colored photographic images of animals, plants, foods and everyday objects. Images were resized to 600 × 600 pixels and presented against a gray background (RGB value: 75, 75, 75). Images were presented on a laptop computer and subtended approximately the central 6° × 5° of visual angle (horizontal × vertical).

Images were divided into equal-size lists depending on the study. Lists were matched for conceptual category membership, lexical decision times on the names and log HAL frequency determined by the English Lexicon Project database (Balota et al., 2007). For the study with epilepsy patients, the stimuli were divided into five lists of 40 items each. For the study with stroke patients, the stimuli were divided into four lists of 50 items each.

Lists were counterbalanced between participants, and for both studies each participant saw all but one of their respective lists of stimuli.

2.2.1. Epilepsy Patient Task Design

Each patient was given the object naming task twice, once about two to three weeks before resection (Pre-resection), and once about three months after resective surgery (Post-resection). Mean (SD) time between task sessions was 112 (17.3) days (range: 86 to 168). All runs of the task consisted of 40 stimulus trials presented in a pseudorandom order. Each trial started with a black fixation cross in the center of the screen for 500 msec, followed by the stimulus picture image to be named for 300 msec. There was a stimulus offset of a blank screen for 2700 msec, and then a randomized jitter of 1000 msec, 2000 msec, or 3000 msec for an average interstimulus interval (ISI) of 4700 msec. Participants were instructed to name each image aloud as quickly and accurately as possible, with correct performance and error responses notated by the experimenter. Responses were recorded by the built-in microphone on the laptop computer and response time was marked according to voice onset relative to stimulus onset.

For the Pre-resection task session, the patient was exposed to one novel list (e.g. List A) of 40 items by performing one run of the task (training). After a break during which the patient participated in other behavioral studies that had no overlap in stimuli or verbal responses (approximately 30–60 minutes), the patient performed two runs of the task (testing). These two runs contained the 40 items from the previously named list (List A) (at a ~30–60 minute delay; for ease of description, this will be referred to as the ‘30-minute delay’ condition) and 40 novel items (e.g. List B). Items from the two lists were intermixed in a pseudorandom order and administered across the two runs of the task. For the Post-resection task session, the patient was first exposed to one novel list (e.g. List C) of 40 items in pseudorandom order for one run (training). After a break, the patient performed three runs of the task (testing). The three runs contained the 40 items that were previously named during training (List C) (30-minute delay condition), 40 items that were previously named during the Pre-resection session (List A) (3-month delay condition), and 40 novel items (e.g. List D). Items from the three lists were intermixed in pseudorandom order and administered over the three runs of the task. See Supplementary Figure 1 for a graphical depiction of this design.

Four patients were given an earlier version of the task design with 50 items per list but all other aspects of the task the same. The four patients were equally divided between the four resection groups (i.e. one patient in each group). The number of items per list was dropped to 40 to reduce the task load for the patients while still allowing for measurable repetition priming.

2.2.2. Stroke Patient Task Design

Patients were given an object naming task in one session. All runs of the task consisted of 50 stimulus trials presented in a pseudorandom order. The instructions, timing of stimulus events, and scoring of responses was the same as described above for Epilepsy patient Task Design.

The first two runs of the task (training runs) consisted of two lists of stimuli (e.g. List A and B) intermixed and presented in pseudorandom order, 50 items per run. The patients then spent approximately 30–60 minutes in an MRI while anatomical and Diffusion Tensor Imaging images were acquired, not performing any behavioral task. After the MRI, the patient performed three more runs of the task (testing runs). During testing runs, the patient saw stimuli from three different lists – the same exact pictures as previously seen in the training runs (List A) (30-minute delay condition), a different picture of the same stimulus item seen during the training runs (List B) (“Exemplar repetition” condition), and Novel items not seen previously (e.g. List C). Items from the three lists were equally distributed between the three runs in a pseudorandom order. The Exemplar repetition condition was not analyzed for the current study.

2.3. Data Analysis

Every response to each individual item for each patient was graded as correct or an error. If an error was made, the type of error was recorded – Omission (no response) or Commission (name of a different object than the presented stimulus): Perseveration (name of a different object than the presented stimulus that was seen previously during the testing session), Different response (a different name is given to an object that was previously seen in the testing session). Responses that were unclear (background or other microphone interference makes the patient’s response onset indeterminable) or had outlying response times (response time is more than 3.5 standard deviations from the mean testing response for the run in which it occurred) – were noted and removed from analysis. Response times to correct naming trials were tabulated for each stimulus condition in the testing runs. For the Pre-resection epilepsy patients, this is the 30-minute delay and Novel condition, and for the Post-resection this is 30-minute delay, 3-month delay, and Novel condition. For the stroke patients this is the 30- minute delay and Novel condition.

In order to adjust repetition priming magnitudes for overall differences in response time (across patients), priming magnitudes were calculated for each patient using effect size (Cohen’s d), the difference in means between repeated and novel conditions divided by the pooled standard deviation (e.g. Gotts et al., 2021, 2015). Other measures of priming strength such as proportional priming [e.g. (Novel-Repeated)/((Novel+Repeated)/2)] that do not explicitly take account of the response time variability in each condition yield similar results in the current study (see Supplementary Tables 1 and 2).

2.4. Statistical Analyses

Statistical analyses were conducted using mixed-effects ANOVAs and t-tests (paired, one- and two-sample t-tests). Of primary importance in the current study are the priming analyses of response time post-resection. Analyses of the effect of group on Neuropsychological tests pre- and post-surgery and analyses of priming magnitudes pre-surgery are of secondary importance but are critical for interpretation of the post-resection effects of group on priming magnitudes. Multiple comparisons performed on the same data (e.g. response time data in the post-resection testing session) have been corrected using False Discovery Rate (e.g. Benjamini and Hochberg, 1995; Genovese et al., 2002), with the p-value thresholds corresponding to q≤.05 and the included tests provided in Supplementary Table 1. Follow-up tests of potential confounds were conducted using ANCOVA with nuisance variables, with an alpha level of .05 used for significance.

2.5. Lesion Masks

A high-resolution T1-weighted anatomical MPRAGE image was obtained prior to resection and three months after resection (239 sagittal slices, 0.8 mm slice thickness, field of view = 24 cm) for each epilepsy patient. For each stroke patient, an anatomical MPRAGE image was obtained between the training and testing portions of the object-naming task (174 sagittal slices, 1.0 mm slice thickness, field of view = 26 cm).

Creation of a lesion mask for each epilepsy patient involved an automated process using AFNI (Cox, 1996). After removing the skull and non-brain tissue surrounding each image, the image taken after resection was aligned to the image taken prior to resection, allowing only for shifting and rotating of the image. A mask was created with all voxels below a raw image intensity value of 1500, thresholded and removed from the aligned post resection image. A lesion estimate mask was created by subtracting the thresholded, aligned post resection mask from a mask of all voxels in the pre resection image. The lesion estimate mask was visually confirmed and manually edited to remove any extra voxels included outside the lesion location.

For the stroke patients, lesion masks were drawn manually by researchers trained to recognize lesional cortical and subcortical structures in T1-weighed anatomical images. Each lesion mask was confirmed by a second, independent trained researcher.

The final lesion mask for each patient was transformed to standardized anatomical space (Talairach-Tournoux) at a resolution of 3 mm³ isotropic voxel size. An ROI of left lateral frontal cortex, for which voxels showed a correlation between “repetition suppression” and repetition priming was calculated from Gotts et al. (2021) (N=60 healthy control participants). A voxelwise statistical threshold was chosen that was most inclusive yet still stringently controlled Type I Family Wise Error below P<.05 when corrected by cluster size for multiple comparisons (P<.002, see Cox et al., 2017; Eklund et al., 2016 for discussion). The ROI was transformed to the same anatomical space as the patient lesion masks and resampled to the same voxel size. An overlap mask was created for each frontal resection or stroke patient from the patient’s lesion mask and the ROI mask. The number of voxels in the overlap mask were calculated and then divided by 97 (the number of voxels in the ROI mask) to determine the percentage of the priming-related left lateral frontal cortex ROI impacted by lesion.

2.6. Data availability

Behavioral data in terms of response times in each condition are available for each patient, along with a lesion mask in Talairach-Tournoux space at figshare.com.

3. Results

Patients were administered a range of standard neuropsychological tests to characterize their overall cognitive, memory, and language functioning both prior to and following resection surgery (see Figure 1 for resection locations). Considering all 41 patients as a full group, mean patient performance prior to resection was consistently in the low neurotypical range (within one to two standard deviations below the mean) on tests of general intellectual functioning (e.g. WAIS), verbal memory (e.g. WMS/BNE, CVTL), and language functioning, including picture naming (e.g. NAB, Boston Naming Test) (see Table 1). Left Anterior Temporal (N=17), Right Anterior Temporal (N=10), Frontal (N=5), and Other (N=9) groups differed significantly on the Picture Naming performance [F(3,34) = 3.18, P=0.0362]. These differences were mainly due to low scores for the Left Anterior Temporal group, as well as relatively higher scores for the Right Anterior Temporal group [Right vs Left, t(24) = 3.12, P=.0046; Right vs Other, t(16) = 2.14, P=.0479; all other paired comparisons not significant]. Neuropsychological assessment results following resection were qualitatively similar to those pre-resection, however with more significant group differences seen between Left Anterior Temporal (N=13), Right Anterior Temporal (N=9), and Frontal (N=4) (see Supplementary Table 3). Frontal patients were not included in group comparisons when the number of scores available was less than four due to small sample size. Differences in WMS/BNE Immediate [F(2,22) = 3.964, P=.0339], WMS/BNE Delayed [F(2,21) = 4.275, P=.0277], CVLT Short Free [F(1,13) = 11.654, P=.0046] CVLT Short Cued [F(1,13) = 13.593, P=.0027] and Picture Naming [F(2,23) = 7.986, P=.0023] were again due to low scores for the Left Anterior Temporal group and relatively higher scores for the Right Anterior Temporal Group [t ≥ 2.923, P≤.0091 on all pairwise comparisons; pairwise comparisons with Frontal when included in group comparison not significant]. The overall neuropsychological profile shows that all patient groups fell within the neurotypical range, with the three groups performing at similar levels for most tests. As subsequent measures of repetition priming during the picture naming experiments are evaluated only on correct trials, small differences in overall performance levels are not expected to confound the results of interest.

Figure 1. Spatial distribution of removed cortex in resection patients by group.

Removed cortex for resection patients (N=40) displayed on the Colin27 template in Talaraich-Tournoux standardized space. Color bar represents the number of patients with removed cortex at a particular voxel. Patients were grouped by general resection area, with an N of five required to separate groups. One patient in the ‘Other Areas’ group with a resection in the right posterior temporal lobe was not included in the map due to MRI-incompatible implants.

Table 1.

Resection Patient Neuropsychological Evaluation Test Scores

	General Cognitive Functioning				Verbal Memory and Learning					Language Functioning
	WASI-II/EIWA-III/WAIS-IV				WMS-IV Logical Memory		CVLT-II			NAB Language Module
					OR BNE Verbal Memory-Prose
Resection Location Group	Full Scale IQa	Verbal Comprehensiona	Perceptual Reasoninga	Digit Spanb	Immediate Recallb	Delayed Recallb	Total Correctc	Short Delay Free Recalld	Short Delay Cued Recalld	Auditory Comprehensiond	Naming OR Boston Naming Testd
Full Group (n=41)
Mean (SD)	85.55 (14.05)	88.42 (14.56)	88.03 (13.23)	6.76 (2.28)	5.95 (3.41)	6.11 (3.41)	41.63 (8.97)	−1.02 (1.4)	−1.07 (1.29)	35.7 (16.11)	35.13 (14.56)
[Range]	[53–120]	[65–122]	[64–121]	[2–11]	[1–14]	[0–14]	[25–57]	[−4.5–1]	[−4–0.5]	[19–57]	[19–58]
n of Test Scores	38	38	39	38	38	37	30	30	30	27	38
Left Anterior Temporal (n=17)
Mean (SD)	82.56 (12.61)	84.33 (12.24)	87.19 (11.92)	7.29 (1.99)	4.2 (2.76)	4.64 (3.65)	38.17 (7.92)	−1.71 (1.32)	−1.75 (1.1)	34.42 (15.89)	29.44 (13.19)
[Range]	[53–102]	[65–110]	[64–110]	[3–10]	[1–10]	[1–14]	[25–51]	[−4.5–0]	[−3.5–0]	[19–55]	[19–58]
n of Test Scores	16	15	16	17	15	14	12	12	12	12	16
Right Anterior Temporal (n=10)
Mean (SD)	88.5 (11.56)	93.7 (13.29)	86.5 (12.36)	6.67 (1.87)	7.4 (2.88)	7.9 (2.33)	41.86 (8.38)	−0.64 (1.25)	−0.64 (0.75)	35.17 (16.47)	45.8 (12.65)
[Range]	[75–108]	[78–113]	[69–105]	[4–10]	[3–11]	[5–12]	[31–57]	[−3–0.5]	[−1.5–0.5]	[19–57]	[26–58]
n of Test Scores	10	10	10	9	10	10	7	7	7	6	10
Frontal Cortex (n=5)
Mean (SD)	86.25 (23.81)	85 (15.03)	89 (21.58)	6 (2.83)	7 (3.74)	6 (2.94)	37 (9.54)	−1.5 (1.8)	−1.33 (2.36)	19 (0)	36.5 (16.09)
[Range]	[64–120]	[65–98]	[67–121]	[2–8]	[2–11]	[2–9]	[27–46]	[−3.5–0]	[−4–0.5]	[19–19]	[19–58]
n of Test Scores	4	4	5	4	4	4	3	3	3	2	4
Other Areas (n=9)
Mean (SD)	87.5 (15.81)	90.89 (18.87)	91 (12.95)	6.13 (3.09)	6.78 (4.02)	6.44 (3.64)	48.38 (8.16)	−0.13 (1.09)	−0.31 (1.13)	43.14 (16.55)	32.5 (13.63)
[Range]	[74–111]	[74–122]	[75–109]	[2–11]	[1–14]	[0–12]	[35–56]	[−2–1]	[−2.5–0.5]	[19–55]	[19–58]
n of Test Scores	8	9	8	8	9	9	8	8	8	7	8
Groups Comparison
F-value	0.422	0.991	0.196	0.650	2.494	1.9629	2.822	2.748	2.742	1.283	3.181e

^aStandardized Score (100 +/− 15)

^bScaled Score (10 +/− 3)

^cT score (50 +/− 10)

^dz score (0 +/− 1)

^eP = 0.0362

3.1. Repetition priming in picture naming: Pre-resection

Prior to resective surgery, each patient was administered a picture naming task using pictures of common living things and man-made objects, with accuracy and naming response times recorded. Patients first named 40 items, and after an approximately 30-minute delay, they named these same items again (Repeated) randomly intermixed with 40 Novel items. In terms of accuracy, there were no significant main effects of Condition [Novel, Repeated after 30 minutes: F(1,37) = 3.522, P=.0685)] or Group [Left Anterior Temporal, Right Anterior Temporal, Frontal, Other: F(3,37) = 1.098, P=.3621] and no significant Group X Condition interaction [F(3,37) = 0.196, P=.8982]. The patients performed at 79.1% correct overall.

In terms of response time, there was a significant effect of repetition priming across all patients [paired t-test: t(40) = 7.77, P = 1.6252×10⁻⁹, FDR-corrected to q<.05; see Figure 2A], with faster response times to Repeated (mean = 986.9 msec, SD = 227.4 msec) compared to Novel pictures (mean 1105.2 msec, SD = 227.7 msec). Since priming effects tend to be larger for slower response times, priming magnitudes for each patient were converted to Effect Size (Cohen’s d) prior to examining effects of Group on priming to adjust for differences in overall response time (see Gotts et al., 2021, for discussion; see also Supplementary Figure 2). There was no significant effect of Group on priming effect size [F(3,37) = 0.6419, P = .5929], nor did any group differ in a pairwise fashion with any other (P ≥ .1939 for all), and each group exhibited significant repetition priming individually (Figure 2B; one-sample t-tests versus 0: t > 3.45, P ≤ .0087 for all, q≤.05). These priming magnitudes are also comparable to those observed in our recent study of repetition priming in picture naming with neurologically intact control participants having an overlapping range of ages (e.g. Gotts et al., 2021; Supplementary Figure 3).

Figure 2. Resection patient naming task performance prior to resection.

(A) Mean response time to correct trials collapsed across all resection patients (N = 41) during naming task prior to resection. Response time to novel stimuli was significantly slower compared to those repeated after a 30-minute delay. (B) Priming effect size (as expressed by Cohen’s d) for stimuli repeated after a 30-minute delay for patients prior to resection. Differences between the four resection groups are not significant. Each resection group separately had significant priming effect size compared to zero. For presentation of the results in terms of mean response time and proportional priming, see Supplementary Table 2.

These results establish that all patient groups exhibited repetition priming following a 30-minute delay prior to the surgical resection and they did not differ markedly from one another in either overall picture naming performance or in priming magnitudes for the current stimuli.

3.2. Repetition priming in picture naming: Post-resection

Three months following resective surgery, we tested the patients again in picture naming to examine if particular resection locations had a differential impact on repetition priming. Patients again began the testing session by naming 40 items they had not seen before. After a delay of approximately ~30–60 minutes patients named these same items again (30-minute delay), randomly intermixed with a new set of 40 items (Novel) and the 40 items that had served in the Repeated condition in the pre-resection testing session (3-month delay). For analyses of the effect of Group on accuracy and response time in the post-resection session, the “Other” resection patient group was excluded, as this group did not represent any consistent location of cortical removal. However, the “Other” group was retained for analyses across all patients. For accuracy, there was a significant main effect of Group [Left Anterior Temporal, Right Anterior Temporal, Frontal: F(2,29) = 5.772, P=.0078, q=.05], but no significant main effect of Condition [Novel, Repeated 30-minute delay, Repeated 3-month delay: F(2,58) = 1.9126, P=.1569] or Group X Condition interaction [F(4,58) = 1.520, P=.2082]. The main effect of Group was largely due to lower accuracy in the Left (mean = 66.75% correct, SD = 17.38%) compared to the Right Anterior Temporal groups (mean = 85.9% correct, SD = 11.17%) [two-sample t-test: t(25) = 3.191, P=.0038, q<.05; all other paired comparisons non-significant].

In response time, there was a significant main effect of Condition across all patients (N=41) [Novel, Repeated 30-minute delay, Repeated 3-month delay: F(2,74) = 21.922, P=3.3365×10⁻⁸, q<.05; see Figure 3A]. Converting to priming effect sizes (Cohen’s d) in order to adjust for overall differences in response time across patients, there was a significant effect of repetition priming at both the 30-minute delay and the 3-month delay [one-sample t-test versus 0, 30-minute delay: t(40) = 7.794, P=1.5071×10⁻⁹; 3-month delay: t(40) = 2.727, P=.0095, q<.05], as well as greater priming at the 30-minute compared to the 3-month delay [t(40) = 5.646, P=1.4855×10⁻⁶, q<.05; Figure 3C].

Figure 3. Resection patient naming task performance three months after resection.

(A) Mean response time to correct trials collapsed across all resection patients (N=41) during naming task approximately three months after resection. Novel stimuli are represented with a solid black bar, stimuli for which first exposure occurred prior to resection (the three-month delay condition) are represented with the cross-hatch patterned bar and stimuli for which first exposure occurred during the same testing session (30-minute delay condition) are represented with the solid-colored bar. (B) Mean response time to correct trials across stimulus condition types for the different resection groups (“Other” resection locations are shown for reference, but were not included in the Group X Condition priming analyses; see text for description). Bar coloring/texture by condition as in (A), with each resection group shown in a different color. (C) Priming effect size after resection for the two delay conditions collapsed across all resection patients (N=41). The 30-minute delay is represented in the solid-colored bar and the three-month delay is represented by the cross-hatch patterned bar. Both effect sizes were greater than zero, with the 30-minute delay greater than the 3-month delay. (D) Priming effect size after resection for the three groups of patients with consistent resection locations. Bar coloring/texture as in (C), with different colors used for different resection groups as in (B). For presentation of the results in terms of mean response time and proportional priming, see Supplementary Table 2.

The effect of resection location on repetition priming was also evaluated with priming effect sizes (Figure 3D, with raw response times by group shown in Figure 3B for reference). There was a significant main effect of priming Condition [30-minute delay, 3-month delay: F(1,29) = 30.055, P=6.6458×10⁻⁶, q<.05], no significant main effect of Group [Left Anterior Temporal, Right Anterior Temporal, Frontal: F(2,29) = 0.366, P=.6966], but a marginally significant Group X Condition interaction [F(2,29) = 2.592, P=.0921, uncorrected]. When examining each pairwise combination of groups, there was a significant Group X Condition interaction involving the Right Anterior Temporal and Frontal groups [F(1,13) = 8.470, P=.0122, q<.05], with the Left Anterior Temporal group failing to significantly interact with either of the other two groups (P≥.1145 for both). The significant Right Anterior Temporal/Frontal interaction appeared to be driven by the lack of priming at the 3-month delay for the Frontal group [one-sample t-test versus 0: t(4) = 0.341, P=.7502] compared to highly significant priming in the 30-minute delay [t(4) = 7.541, P=.0017, q<.05]. In contrast, the Right Anterior Temporal group exhibited priming at both delays, similar to the results for all patient groups combined [30-minute delay: t(9) = 5.534, P=3.6379×10⁻⁴, q<.05; 3-month delay: t(9) = 2.490, P=.0344, just failing to survive FDR-correction, q=.06]. While not directly involved in this interaction, it is worth pointing out that priming at the 3-month delay, while numerically larger than that observed in the Frontal group, also failed to reach significance in the Left Anterior Temporal group [t(16) = 1.112, P=.2825], with significant priming observed after a 30-minute delay [t(16) = 3.758, P=.0017, q<.05].

3.3. Repetition priming following lateral frontal cortex damage

Tentatively, the current results highlight the potential involvement of frontal cortex in mediating very long-lasting effects of repetition priming (3 months or longer). However, they also potentially conflict with previous neuroimaging results indicating the involvement of frontal cortex for the more typical 30-minute delay (e.g. Dobbins et al., 2004; Ghuman et al., 2008; Gotts et al., 2021; Horner and Henson, 2008; Maccotta and Buckner, 2004; and see Wig et al., 2005 for the effects of left frontal transcranial magnetic stimulation on repetition priming in healthy controls). Previous results have implicated the involvement of the left lateral frontal cortex in the territory of the inferior frontal junction, whereas four of the five frontal resection cases in the current study involve right anteromedial frontal regions (Figure 1), with only one patient’s anterior frontal removal impinging on the left lateral frontal cortex. For this reason, we examined repetition priming following a 30-minute delay in an additional group of 10 recovered aphasics with lesions due to stroke, some patients with damage overlapping the left inferior frontal junction and some with no overlap (Figure 4A; neuropsychological test performance shown in Table 2). As with the resection cases, these patients performed in the low normal range for executive functioning and attentional processing (e.g. TEA), as well as for verbal and visual memory (e.g. CVTL; Doors and People Assessment), and had very mild residual language impairments (scores between 90 and 100 on the WAB) (Table 2). Importantly, in terms of overall picture naming performance, the stroke patients did not differ from the resection cases in accuracy or response time [Stroke versus all resection patients, Accuracy: t(49) = 0.637, P=.5271, Response Time: t(49) = 0.304, P=.7627; Stroke versus frontal resection patients, Accuracy: t(13) = 1.137, P=.2762, Response Time: t(13) = 1.023, P=.3249; see Figure 4B].

Figure 4. Stroke patient naming task performance and the involvement of left lateral frontal cortex in repetition priming.

(A) On top, the lesioned cortex for stroke patients (N=10) displayed on the Colin27 template in Talaraich-Tournoux standardized space. Color bar represents the number of patients with lesioned cortex at a particular voxel. On bottom, the ROI in left lateral frontal cortex associated with priming in prior studies (centered at coordinate −42, +11, +31) is outlined in black over the spatial distribution of lesioned cortex in stroke patients. (B) Mean response time for correct trials across novel stimuli (solid black bar) and those repeated after a 30-minute delay (solid red bar) for stroke patients, with all resection patients after resection (N=41) and the frontal resection group, in particular, shown for reference. Stroke patients’ response times did not significantly differ from resection patients when collapsed across group or when compared to frontal resection patients. (C) Priming effect size for patients with lesioned or resected cortex that overlaps with the left lateral frontal cortex ROI (N=8) compared to patients with no overlap (N=7). The two separate groups are represented by the clear bar graphs, with the priming effect size for individual patients represented by the colored circles. Circles in yellow are resection patients and circles in blue are stroke patients. The percentage next to each individual patient’s circle in the overlap group is the percentage of ROI voxels overlapped by damage. Priming was significant for the No Overlap group and was significantly greater than for the Overlap group. For presentation of the results in terms of mean response time and proportional priming, see Supplementary Table 2.

Table 2.

Stroke Patient Neuropsychological Evaluation Test Scores

	General Cognitive and Executive Functioning									Verbal Memory and Learning										Language Functioning
	Test of Everyday Attentiona									Doors and People Assessmenta							CVLT			Western Aphasia Batteryd
	MSIM	MS2M	ECWD	VEA	VET	ECWR	TS	TSWC	L	VisM	VerM	Rcl	Rcn	VisF	VerF	Total	Totalb	SDFRc	SDCRc	Aphasia Quotient	Language Quotient	Cortical Quotient
Overlap with Left Lateral Frontal ROI
Patient l	8	5	5	7	1	7	9	10	2	12	14	12	14	9	11	14	31	−2	−2	91.8	88.2	89.9
Patient 2	8	10	2	3	1	6	2	3	2	17				11			22	−2.5	−2.5	85	86	87
Patient 3	8	5	3	4	8	8	4	11	13	8	9	8	9	18	5	8	30	−1.5	−2	92.2	93	92.5
Patient 4	6	2	7	7	3	6	5	9	2	10	9	10	9	11	10	9	33	−1.5	−1.5	90.2	86.7	88.4
Patient 5	7	6	8	6	1	6	3	1	6	14	7	11	10	11	11	11	36	−1.5	−2	88.2	92.8	92.1
Patient 6										7	4	5	6	11	10	5				87	91.9	91.2
Patient 7										7	9	7	9	l4	12	7				93.6	89	91.1
No Overlap with Left Lateral Frontal ROI
Patient 8	3	1	3	5	2	7	2	7	4	3	6	3	6	6	3	2	15	−4.5	−5	96.8	97.7	97.5
Patient 9	5	3	12	6	1	5	7	8	4	5	8	6	6	11	4	6	48	−3	0.5	88.8	91.3	91.6
Patient 10	7	3	4	2	1	5	4	5	2	10	3	9	3	11	10	5	32	−2	−1.5	95.6	92.5	94.8
Group
Mean	6.5	4.38	5.5	5	2.25	6.25	4.5	6.75	4.38	9.3	7.67	7.89	8	11.3	8.44	7.44	30.88	−2.31	−2	90.92	90.91	91.61
SD	1.77	2.83	3.34	1.85	2.43	1.04	2.45	3.49	3.78	4.22	3.24	2.93	3.16	3.09	3.43	3.57	9.69	1.03	1.51	3.78	3.5	3

^aScaled Score (10 +/− 3)

^bT Score (50 +/− 10)

^cz score (0 +/− 1)

^dScore is out of 100, with 100 being normal function

MS1M = Map Search 1 Minute; MS2M = Map Search 2 Minute; ECWD = Elevator Counting with Distraction; VEA = Visual Elevator Accuracy; VET = Visual Elevator Timing; ECWR = Elevator Counting with Reversal; TS = Telephone Search; TSWC = Telephone Search while Counting; L = Lottery; VisM = Visual Memory; VerM = Verbal Memory; Rcl = Recall; Rcn = Recognition; VisF = Visual Forgetting; VerF = Verbal Forgetting; SDFR = Short Delay Free Recall; SDCR = Short Delay Cued Recall

For purposes of calculating spatial overlap between previous neuroimaging correlations with repetition priming and patient lesions, we used an independently defined region of interest from our recent study of repetition priming in picture naming for neurotypical participants (N = 60; Gotts et al., 2021), the outline of which is rendered for reference in Figure 4A. Seven of the 10 stroke patients had lesions overlapping the frontal region of interest, with three having no overlap. The performance of these patients was then pooled with the 5 frontal resection cases, also sorted by overlap with the frontal region of interest (one having overlap, and 4 with no overlap). This resulted in a group of eight patients with damage overlapping the left frontal region of interest (Overlap group), and seven patients with no overlap (No Overlap group).

A comparison of priming effect sizes at the 30-minute delay for patients with versus without damage overlapping the critical left frontal region of interest is shown in Figure 4C by damage etiology (stroke or resection). The Overlap group showed significantly reduced repetition priming compared to the No Overlap group [two-sample t-test: t(13) = 2.740, P=.0169, q=.05], with the Overlap group failing to show significant priming [one-sample t-test: t(7) = 1.803, P=.1144] and the No Overlap group showing highly significant priming [t(6) = 11.966, P=2.0643×10⁻⁵, q<.05]. These results did not appear to be easily explained by etiology. An ANCOVA including etiology (resection, stroke) as a categorical nuisance covariate continued to demonstrate a significant effect of Overlap [F(1,12) = 4.88, P=.0473], with no effect of etiology [F(1,12) = 0.06, P=.8105] and no interaction with group when modeling separate covariate slopes for the individual groups [F(1,11) = 0.56, P = .4695].

Taken together, these results provide convergent evidence that left lateral frontal cortex is necessary for intact repetition priming following approximately 30–60 minutes. In contrast, damage to more anterior and medial frontal regions may be more critical for very long-lasting effects of repetition priming at delays of 3 months or more.

3.4. Relationship of repetition priming to explicit memory

In the current study, repetition of stimuli was implicit to the task being performed. However, participants were likely aware that repeated stimuli had been presented previously, especially in the 30-minute delay conditions. It remains possible that explicit memory, as measured by free recall or recognition memory, and the priming effects measured here (a typical measure of implicit memory) share some form of interdependence in the patients (for discussion, see Duff et al., 2020; Greenberg and Verfaellie, 2010; Henson and Gagnepain, 2010; Voss and Paller, 2008). We therefore utilized the pre- and post-resection neuropsychological assessment scores on explicit memory performance (free recall measures from the CVTL, WMS, BNE) to examine possible contributions of explicit memory to the priming effects observed post-resection. As the stroke patients also had measures on the CVTL, they were also included for correlations involving the CVTL (for the 30-minute delay). WMS or BNE recall measures were only available for the resection patients.

On the WMS Logical Memory Immediate Recall and Delayed Recall, the BNE Verbal Memory – Prose Immediate Recall and Delayed Recall, the CVLT Total Correct, Short Delay Free Recall and Cued Recall, there were no significant correlations with priming effect size (Cohen’s d) across patients in either the 30-minute or 3-month delay conditions [Pre-resection WMS, BNE, CVTL scores with Post-resection priming effect sizes: r(28 or more) < |.26| for all, P≥.16 for all; Post-resection WMS, BNE, CVTL scores with Post-resection priming effect sizes: r(22 or more) < |.19| for all, P≥.39 for all]. These relationships failed to be observed despite the fact that sufficient variability was present in the WMS scores, BNE scores, CVTL scores, and priming effect sizes to detect significant differences among the patient groups (see Table 1, Supplementary Table 3, and Figure 3).

As a second check, we examined the dependence on explicit memory of the observed Group (Frontal, R ATL) X Condition (30-minute delay, 3-month delay) interaction in priming magnitude. The WMS-IV Logical Memory was the most common test of memory present for these patients in the post-resection testing session, along with the BNE (Verbal Memory – Prose), with one patient in each group missing scores. With a scaled score for delayed recall serving as a nuisance covariate in ANCOVA for each patient, the interaction remained significant and was robust to assumptions about the values of the two missing scores (P≤.0248 for all options examined), with little dependence observed on the explicit memory measures (see Supplementary Figure 4).

Taken together, the current results indicate that the inter-patient variability in priming magnitudes is not well explained by inter-patient variability in explicit memory ability.

4. Discussion

In the current study, we have examined the magnitude of repetition priming, an important form of implicit memory, in picture naming both before and after resection surgery for the treatment of epilepsy in 41 patients with a variety of seizure foci. Patients were in the low neurotypical range in cognitive functioning both before and after surgery, and significant repetition priming was observed over all patients when considered as a single group at both typical 30-minute and very long-term (3-month) delays. Given the overlap of the resection locations in the current study with activations in neuroimaging studies of the same picture naming task using the same stimuli (e.g. Gilmore et al., 2019; Gotts et al., 2021; see also Kan and Thompson-Schill, 2004), the relative sparing of priming is likely mediated by interactions among numerous distributed brain regions. Nevertheless, the current results also show that priming can be compromised under damage to frontal brain regions, with distinct effects at different frontal locations and different temporal delays. Removal of antero-medial frontal regions appears to eliminate repetition priming of items named 3-months prior to resection surgery, but spares priming at the shorter, 30-minute delay. Indeed, patients with a frontal resection had priming effects post-surgery at the 30-minute delay that were numerically larger than all other groups. In contrast, damage to lateral frontal cortex in the territory of the inferior frontal junction significantly impairs priming at a 30-minute delay, as shown in the combined analyses of the five frontal resection cases and 10 recovered stroke aphasics. While it was not possible to investigate the role of left lateral frontal cortex in mediating much longer-term priming due to the absence of enough patients with this resection location, priming magnitudes tend to show reduced magnitude over time (e.g. Cave, 1997; Henson et al., 2004; van Turennout et al., 2003). We therefore think it highly likely that longer-term priming would be relatively impaired in such patients, as well.

Importantly, reduced priming magnitudes at a 30-minute delay following left lateral frontal cortex damage support the interpretation of Wig et al. (2005) and others (e.g. Horner, 2012) that left lateral frontal cortex is critical for intact repetition priming. As stated earlier, the potential for poly-synaptic spread in transcranial magnetic stimulation (e.g. Bergmann and Hartwigsen, 2021; Cowey, 2005; Logothetis et al., 2010) cautions against using this method as the sole source of assessing causality. The convergent evidence from neuropsychology provided here helps to bolster the conclusion that left lateral frontal cortex is indeed necessary for intact priming (see also Martin and Gotts, 2005, for discussion). Swick (1998) studied repetition priming in lexical decision with 11 frontal patients with damage due to stroke but failed to find an impairment in repetition priming. The delays used in this study were much shorter (60 seconds or less) than those used in the current study, and the patient lesions were concentrated ventrally to the inferior frontal junction. It is possible that differences in task, lesion location, and delay contributed to the discrepant findings between the studies.

Ghuman et al. (2008) demonstrated in magnetoencephalography an early and transient synchronization of activity between left lateral frontal cortex and the left fusiform gyrus in the low beta frequency range (12–14 Hz) that was correlated with priming magnitude at this same delay. It is possible that this “synchrony” effect underlies the left lateral frontal contribution to priming (see also Gotts et al., 2012). However, there are also other possibilities, including the retrieval of recent stimulus-response associations through cortical-striatal loops (e.g. Gotts et al., 2021; Henson et al., 2014), as well as increased top-town suppression from frontal to ventral temporal activity in predictive coding (e.g. Ewbank and Henson, 2012; Friston, 2012, 2005; Korzeniewska et al., 2020). It should also be noted that these possibilities are not mutually exclusive. We found no evidence in the current study that priming magnitudes at either delay depended on or interacted with episodic memory abilities.

The observation of impaired longer-term repetition priming (3-month delay) following anteromedial frontal cortex removal is novel to the current study. The finding is also unique in that the previous naming experience with the corresponding stimuli was prior to the surgery. This suggests that while this cortex does not appear to be necessary for within-session priming effects (a 30-minute delay), it may have been engaged and critical for processing at the time of that initial experience, with a greater role in mediating much longer-term priming effects. What is needed in order to further clarify the role of antero-medial frontal cortex in repetition priming at long lags are additional neuroimaging studies at these longer lags. To date, the longest lag used in neuroimaging studies of priming has been 3 days (e.g. van Turennout et al., 2000; 2003), far short of the 3 months used here. The current results would seem to predict that the neural correlates of repetition priming should shift locations within frontal cortex from lateral to antero-medial over this time window.

4.1. Limitations of the current study

The total number of patients examined in the current study is moderate to large for studies of this type (41 resection patients and 10 recovered stroke aphasics). However, the number of patients in each selected anatomical resection group is much smaller, with 17 patients in the Left Anterior Temporal group, 10 patients in the Right Anterior Temporal group and only five patients in the Frontal group. Caution is warranted in generalizing from a group of only five patients, although the reason effects were observed in this group is that the variability among the patients at both the 30-minute and 3-month priming delays was found to be low with relatively consistent priming levels across the group (Figures 3D and 4B; see also Supplementary Figure 2). It will be important for future studies to examine very long-lasting priming effects in additional patients, perhaps with other etiologies than those examined here.

The priming pattern in the left anterior temporal lobe patients failed to interact with either the right anterior temporal or frontal groups, although this may be due to a lack of statistical power. Like the frontal patients, this group had significant priming after a 30-minute delay but failed to exhibit significant priming after 3 months. These patients also had significant impairments in picture naming and in explicit memory following resective surgery (Supplementary Table 3). Our failure to find significant differences in priming magnitude following left anterior temporal damage should be revisited in the future with a larger number of patients.

There was no independent neurotypical control group in the current study. Instead, patients served as their own controls pre- and post-resection surgery. The priming magnitudes in terms of effect sizes observed by the epilepsy patients here are also comparable to those seen in neurotypical controls of approximately the same ages and when using these same stimuli (e.g. Gotts et al., 2021). Comparisons were also made across patients with different resection or stroke locations, with the same etiology present across the different compared conditions (e.g. Frontal versus Right Anterior Temporal resection; Stroke with or without lesion overlapping the left lateral frontal region of interest). Long repetition priming lags have been examined behaviorally in both patients (e.g. 1-week, amnesia: Cave and Squire, 1992) as well as in neurotypical controls (e.g. 6 to 48 weeks: Cave, 1997). However, detailed comparisons of effect sizes between patients and controls at these longer lags have not been conducted, and our understanding of typical versus impaired levels would benefit from additional studies in neurotypical controls.

Our use of both epilepsy and recovered stroke aphasics has the potential to confound priming deficit with etiology. Biases in the two groups for lesion location are difficult to prevent, since resection surgery to remove the left frontal cortex is usually avoided to preserve language functioning, and this lesion location will be more likely in stroke aphasics. Lesion sizes are also typically larger in stroke aphasics, although analysis of the relationship between lesion size and priming magnitude suggests that the size of lesion was not determinative in this study (see analyses in Supplementary Text). Based on the neuropsychological assessments, both patient groups performed in the low neurotypical range on tests of executive functioning and attentional abilities, as well as on tests of verbal learning and explicit memory. Both patient groups also have conditions that persist over years (epilepsy versus permanent lesions resulting from stroke). Our analyses also revealed no overall difference in naming performance or priming levels by etiology. However, it is possible that our testing was not sensitive enough to detect such differences. These potential issues highlight the importance of obtaining converging evidence from neuroimaging in neurotypical controls, demonstrating a similar shift in the neural correlates of priming across delays.

Supplementary Material

1

Supplementary Figure 1 Task design for patient groups. Each box represents a unit of time, either one run of the task or the 30 to 60 minute delay between training and testing. Within each box is listed the overall run number within the task session, the number of trials in the run, and the list identifier for the stimuli presented during the run. A fraction before a list identifier means that only that fractional portion of the list was presented during that run. At the end of each row of boxes, each stimulus list presented during the task session is equated to the stimulus condition used in the analyses. The top row shows the design for epilepsy patients prior to resection, the middle row shows the design for epilepsy patients three months after resection, and the bottom row is the task design for the single visit with the stroke patients. For the stroke patients, the exemplar condition (equated with List B in this figure) was dropped from further analysis because it was used in a prior task design and there was no equivalent condition for the epilepsy patients.

2

Supplementary Figure 2 Priming effect sizes are not systematically related to mean accuracy or response time. (top panel) Overall accuracy (proportion correct) in the picture naming task (post-resection testing session) is strongly related to overall response time (in milliseconds) across patients; slower reaction times are associated with lower accuracies. Given that raw priming effects tend to be larger for slower mean response times (and that mean response time varies by group), it was important to convert priming effects to Effect Size (Cohen’s d: the difference in means divided by the pooled standard deviation). After the conversion to Effect Size, there is no significant correlation (Spearman) of priming magnitude with mean accuracy or mean reaction time for effects at the 30-minute delay (middle panels) or the 3-month delay (bottom panels). These results indicate that the reported priming effects are not strongly biased by differences in accuracy or overall response time between the patient groups. Scatterplots are shown for each correlation analysis using color to indicate the patient group (either resection location or stroke).

3

Supplementary Figure 3 Priming effect sizes pre-resesction are comparable to those observed in neurotypical controls. Data from Figure 2B (Pre-resection) is shown on the left for reference. Data from N=58 neurologically intact controls (23 male, 35 female) with overlapping ages (mean = 23.8, SD = 2.88, range: 19–38) from Gotts et al. (2021, Communications Biology) using the same stimulus set and delay (30 minutes) is shown on the right. None of the individual group Pre-resection priming magnitudes differed significantly from the control values (P>.16 for all comparisons). Error bars represent standard error of the mean. Control data available on Figshare: [https://figshare.com/articles/dataset/Priming_measures_for_overt_and_covert_naming_participants_in_Gotts_Milleville_Martin_2021_Communications_Biology_/17088980/1].

4

Supplementary Figure 4 Interaction of priming condition by group is unrelated to delayed recall ability. The interaction of group (Frontal, R ATL) X delay condition (30-minute, 3-month) was examined in relation to a measure of explicit memory from the post-resection period, scaled scores from WMS-IV Logical Memory Delayed Recall or BNS Verbal Memory – Prose Long-Delay Free Recall (both measures of long-term explicit memory for narratives). Two patients (one Frontal and one R ATL) had missing scores, which were estimated through mean replacement (A), K-nearest-neighbor imputation using the pre-resection neuropsychological scores for which these two patients had values (B) (with choices of K from 1–5) (as implemented in Matlab by Khan, 2021; for discussion, see Kim et al., 2004; Mills et al., 2019), and exclusion of the cases with missing scores (C). ANCOVAs were run using the difference between priming effect size in the 30-minute and 3-month delay conditions as the dependent variable and WMS-IV/BNS Delayed Recall scaled score as the nuisance covariate. Scatterplots in A-C show that the assumption of ANCOVA of the same slope of the covariate on each group holds well for these data (Frontal patients shown as yellow circles and R ATL patients shown as green squares), and the effect of group appears to be additive along the y-axis with slopes near zero. Further analyses determined that significant effects of group did not depend on assumptions of missing value replacement, with any choice of scaled score from 2 though 18 (scaled scores have a mean of 10 and SD of 3) for the missing values yielding significant results [F(1,12) > 6.5, P≤.0248 for all]. Cases with missing covariate values in A and B are highlighted by circles with dashed lines, and yellow and green dashed lines represent the best-fit lines of each individual group on the covariate and dependent measure.

Highlights.

• Effect of surgical resection/stroke on long-term repetition priming investigated

• Priming selectively impaired at 3 months with anteromedial frontal removal

• Shorter-term priming (30-minutes) impaired instead with lateral frontal damage

• Priming effects unrelated to individual variability in explicit memory or etiology

Abbreviations:

CVTL

California Verbal Learning Test

BNE

Batería Neuropsicológica en Español

EIWA

Escala de Inteligenica de Wechsler para Adultos

IQ

Intelligence Quotient

MCA

Middle Cerebral Artery

NAB

Neuropsychological Assessment Battery

msec

milliseconds

MPRAGE

magnetization-prepared rapid gradient-echo

ROI

Region of Interest

SD

Standard Deviation

TEA

Test of Everyday Attention

WAB

Western Aphasia Battery

WASI

Wechsler Abbreviated Scale of Intelligence

WAIS

Wechsler Adult Intelligence Scale

WMS

Wechsler Memory Scale