Behavioral response times were averaged across 12 subjects for each condition. The average behavioral response latencies and standard deviations for 12 subjects were: New-Hit words 1273.8 (117.2); Old-Hit words 1103.6 (104); Old-Miss words 1284.1ms (162); and New-Miss words 1278.2 (106). Mean reaction time differed significantly between New-Hit and Old-Hit trials t(11) = 3.69, p < .004; between Old-Miss and Old-Hit trials t(11) = 2.39, p < .036 ; and between New-Miss and Old-Hit trials t(11) = 3.65, p < .004. Furthermore, subjects’ accuracy was good and the average hit (.754) and false-alarm (.353) rates for Old words demonstrated significant differences t(11) = 12.65, p < .001.
When comparing the average number of words remembered vs. those that were forgotten, subject responses were not largely biased based on frequency (Kucera, 1967
), number of letters or imagery (Coltheart, 1981
). On average there was no difference in frequency for words (frequency ratings used when available) selected as old (45.96 ± 2.0, mean ± std. err.) and those selected as new (41.90 ± 1.9), t
(22) = 1.25, p
< .16. There was a slight difference in the number of letters in words selected as old (6.28 ± .02) vs. new (6.13 ± .04), t
(22) = 3.09, p
< .007. Imagery ratings, when available, were found to be slightly higher for words selected as new (511.44 ± 3.1) than those selected as old (489.38 ± 2.2), t
(22) = 5.86, p
Overall brain response pattern during word retrieval
Both novel and repeated trial types demonstrated a highly similar posterior to anterior sequence of brain response consistent with classical models of language processing (Benson, 1979
; Geschwind, 1979
). This characteristic pattern is exemplified by response to the Old-Hit condition shown in .
Statistical parameteric maps of spatiotemporal response patterns during delayed recognition (Old-Hit trials)
The initial visual response peaks bilaterally at ~95ms post-stimulus in the occipital pole and calcarine sulcus (primary visual cortex). Response then spread rapidly to anteroventral occipital cortex corresponding approximately to Brodmann area 19 (~BA 19) where by ~140ms it was strongest within the language dominant, left hemisphere. This early response extended into left posterior fusiform and slightly more anterior occipitotemporal areas (~BA 37/19) previously implicated in the encoding of letter-strings and other word-like stimuli (Nobre et al., 1994
). Response also occurred within the left posterior cingulate area (including precuneus and retrosplenial cortices) at this time. By ~240ms response continues within ventral temporal regions and is maximal in posterior language areas, including Wernicke’s area (~BA 22) and adjacent temporo-parietal cortex both implicated in lexico-phonemic/phonological processing (Rumsey et al., 1997
; Paulesu et al., 2000
; Price, 2000
) and occipitotemporal (ventral ~BA 37/19) cortex implicated in the processing of visual form attributes of words (Tarkiainen et al., 1999
; Dhond et al., 2001
; Tarkiainen et al., 2002
; Marinkovic et al., 2003
). At this time, response engaged the entire posterior half of the temporal lobe and also began within bilateral anterior cingulate cortex and medial aspects of the post-central gyrus. Response also began within anterior temporal and prefrontal regions but is not maximal until longer latencies.
The dominant focus
of cortical response continued to move forward, and by ~340ms encompassed anterotemporal regions (~BA 38, and anterior 20/21) associated with lexico-semantic processing (Hodges et al., 1992
; Damasio et al., 1996
; Mummery et al., 2000
). Prefrontal regions were also active at this time, and their response increased over the next 100ms, peaking at ~400ms and strongest within the left hemisphere (including ~BA 11, 47, 45, 44, 10). The cortical distribution and timing of these responses were similar to those of the N400m as measured by MEG (Dhond et al., 2001
; Halgren et al., 2002
; Marinkovic et al., 2003
) and intracranial EEG (Smith et al., 1986
; Halgren et al., 1994b
; Halgren et al., 1994a
; Guillem et al., 1995
; McCarthy et al., 1995
). Other MEG studies utilizing single or a few dipoles have found context sensitive N400-like components at similar latencies localizing within temporal and/or frontal regions (Simos et al., 1997
; Helenius et al., 1998
; Penney et al., 2003
; Puregger et al., 2003
; Pylkkanen and Marantz, 2003
). At this time, there was also strong bilateral recruitment of cingulate cortex, which may play a role in attention and working memory as well as response choice (Bush et al., 2000
; Braver and Barch, 2002
; Cabeza et al., 2003
; Wang et al., in press
). Response also occurred within precuneus and retrospenial areas. Fronto-temporal response continued until the end of the epoch, with lateral prefrontal areas showing their greatest response at longer latencies (~800ms onward). These areas included classical Broca’s area (BA 44) (Benson, 1979
) and neighboring regions (~BA 9/46/45), especially of the left, as well as anterior/dorsal (~BA 10 and 9) and ventral prefrontal cortices (~BA 47/11) on the right, often activated during memory retrieval tasks (Cabeza and Nyberg, 2000
New-Hit vs. Old-Hit: Repetition with recognition (the New/Old effect) ()
In order to assess the effect of item repetition during successful recognition, we contrasted New-Hits with Old-Hits ().
Differential response between New-Hit and Old-Hit conditions first occurred within the right, anteromedial temporal region, corresponding to entorhinal and perirhinal cortices at ~310ms followed by left anterior insular cortex at ~425ms. There were also differences within the posterior cingulate area including the precuneus and retrosplenial cortex. By ~455ms, response spread anteriorly to left ventral prefrontal cortex (~BA 11, 47). Differences began at ~490ms, in more lateral prefrontal regions (~BA 45, 46) including Broca’s area (~BA 44) and also within left anterior/dorsal prefrontal cortex. However, these responses only became strong and sustained at longer latencies (>600ms). All responses (between 300–500ms) were greater for correctly classified novel items, and all continued on through longer latencies (compare , New-Hit vs. Old-Hit). In the right hemisphere, differences within similar lateral and ventral prefrontal as well as dorsal prefrontal regions were significant from ~600ms onward. , New-Hit vs. Old-Hit, shows that the greatest average difference between 500–700ms occurred within the right anteromedial temporal cortex (entorhinal/perirhinal and parahippocampal areas), and left ventral prefrontal cortex. Starting at ~660ms, but most significant from ~700–900ms, the differential response in the left hemisphere shifted posteriorly to include occipitotemporal regions encompassing the fusiform and parahippocampal gyri. There were also small differences within lateral parietal regions at longer latencies (~600–900ms). All responses were greater for New-Hit words with the exception of the response within Broca’s area, which was greater to repeated words. Responses within all of the above regions gradually peaked at ~815ms post-stimulus, and then continued until the end of epoch (1000ms.)
Overall these results suggest that successful recognition depends critically on the semantic circuits in the anterior temporal (~BA 38/20) and ventral prefrontal (~BA 47/11) cortices. Furthermore, at these longer latencies, the differences increasingly involved bilateral anterior/dorsal (~BA 9,10) and lateral prefrontal (~BA 44/45) perhaps supporting directive retrieval processes (Buckner and Koutstaal, 1998
; Wheeler and Buckner, 2003
Significant cortical responses for successful recognition following a long delay (New-Hits vs. Old-Hits)
New-Hit vs. Old-Miss: Repetition in the absence of recognition
To determine if the responses to delayed word repetition noted above are specific to successful retrieval New-Hits were contrasted with Old-Miss trials (). In both conditions, the words were classified as new, but in the Old-Miss trials they were in fact repeated words.
Effects of Actual vs. Perceived Repetition/Novelty: A: New-Hit vs. Old-Miss
The differences began slightly later (at ~530ms post-stimulus) than those described above where Old words were correctly recognized. The most significant effects between 500 and 700ms were located within the right anteromedial temporal and bilateral posterior cingulate areas, with additional responses in bilateral posterior parahippocampal and fusiform gyri activity. These differential responses increased and continued at longer latencies (700–900ms) when there was also response within ventral prefrontal regions bilaterally but more significant in the left hemisphere.
These results contrast with the findings above showing that, when Old words were not correctly recognized, they produced less response in many of the same areas as when they were correctly recognized, including bilateral anteromedial temporal, ventral prefrontal, and posterior cingulate areas. However, these decreases began later and were less significant. Furthermore, repetition in the absence of recognition did not alter response within left lateral prefrontal cortex (~BA 44/45) suggesting that this area may participate in accurate explicit recognition. Overall these data suggest that beginning at an early latency, sustained differential response within bilateral prefrontal regions may be necessary for successful retrieval.
Old-Hit vs. Old-Miss: Repetition without Recognition ()
A similar response pattern was observed when contrasting activity evoked by repeated words that were recognized vs. those that were not. Again, the main areas responding differentially were ventromedial temporal, ventral and lateral prefrontal and posterior cingulate areas. Of these, the first significant differences occurred within the right anterior middle prefrontal cortex at ~450ms post-stimulus. Differential response then spread to the right frontal pole and inferior prefrontal regions by ~510ms and began at ~540ms within the right anteromedial temporal cortex continuing through longer latencies (~800ms). At ~ 650ms small differences occurred within the left inferior prefrontal cortex (~BA 47/11, 44) and temporal pole, then within bilateral posterior cingulate cortex at ~600ms (left > right) and left posterior fusiform and parahippocampal gyri at ~740ms post-stimulus.
Thus, correctly recognized repeated words (Old-Hit) demonstrate increased response within Broca’s area, when they were contrasted with words perceived as new, regardless of whether the words are actually new () or had actually been seen before (the current contrast). This again suggests that increased response within Broca’s area may to some extent be associated with successful retrieval of words or explicit recognition. Similarly, increased response within the right anteromedial temporal cortex may signal the eventual categorization of an item as novel (even if it is not).
New-Hit vs. New-Miss: Perceived Repetition in the absence of Actual Repetition
To gain more insight into the brain areas that were differentially affected by perceived ‘oldness’, novel words misperceived as ‘old’ (New-Miss), were subtracted from correctly classified novel trials (New-Hit) as seen in . Response was greater to trials correctly classified as new (New-Hit trials) from ~395ms onward in the right anteromedial temporal cortex. As seen earlier, larger response power within this area also occurred for words categorized as “new” (even if they were not); thus, sustained response within this area may indicate the perception of a stimulus as novel, though the strongest response occurs when judgments are accurate. By ~500ms there was also differential response within the right posterior cingulate area, and right dorsal prefrontal regions by ~600ms. Differences began within the ventral prefrontal in the right at ~625ms and left from ~670ms onward. The prefrontal differences peaked (bilaterally) between 800–850ms. There were also small differences within left medial and polar temporal cortex, as well as strong responses in the right anterior, medial and lateral temporal lobe from ~700ms onward.
The overall network seen in this contrast is highly similar to that observed in the previous contrasts. Relatively little differential response was seen within ventral occipitotemporal or left lateral prefrontal regions suggesting that these areas may be slightly more sensitive to actual repetition or its recognition. Conversely, relatively more significant differences occurred in right prefrontal and anteromedial temporal areas, suggesting that the perception of repetition may be sufficient to modulate response within these areas.
Old-Hit vs. New-Miss: Actual Novelty in the absence of Perceived Novelty
Further analysis of the effects of actual repetition were assessed by considering the effects of words incorrectly vs. correctly perceived as ‘old’ (.) Significant differences occurred mainly at longer latencies >700ms, and were smaller than those of the preceding contrasts. Response peaked ~915ms within left parahippocampal and ventral occipitotemporal regions as well as right anteromedial temporal cortex. Differential response began in right ventral and polar prefrontal regions at ~800ms. Differences were strongest within the left ventral prefrontal cortex from ~900ms onward. All of these responses were greater for the New-Miss condition. Again, the overall network observed was similar to that found for the other contrasts, but was late and weak. Preserved differential response in the right anteroventral temporal, and left ventral occipitotemporal cortices suggest that these areas may weakly respond to repetition even when it is not sufficient to guide behavior.
Comparisons of High and Low confidence Trials
The comparison of high and low confidence trials for novel words demonstrated increased response to high-confidence trials within regions of the right anteromedial temporal and parahippocampal cortices from ~300–900ms post-stimulus (). This association of confidence in novelty judgments with right medial temporal response is consistent with the notion that novelty detection is based on a signal from that area (Tulving et al., 1996
). For repeated words the comparison of “high” confidence judgments with “low” confidence trials yielded no sustained significant differences.
Comparison of novel high vs. low confidence trials