The onset latencies associated with neural novelty responses in recognition memory studies differ largely between nonhuman primates (~70–80 ms) and humans (~150–200 ms). Although this large species difference in the speed with which neural novelty signals emerge could potentially be explained by brain size, there is also an important experimental factor that so far has not been fully considered. Nonhuman primates are usually motivated to discriminate novel and familiar stimuli by rewarding them for either detecting the novel item or detecting the familiar item, and in some studies, responses to both novel and familiar items are rewarded 
. In human recognition memory studies, on the other hand, reward is not used to motivate the detection of novel or familiar items. Remarkably, the possibility that the timing of neural novelty signals might be affected if the discrimination of novel and familiar items is rewarded has not yet been tested. Indeed, novelty processing engages neurotransmitter systems that play an important role in the regulation of motivational aspects of behavior, most notably dopaminergic circuitry [9–12]
. Furthermore, reward motivation can energize behavior 
, leading to decreased response times 
and increased response vigor [15, 16]
In two recognition memory experiments, we used magnetoencephalography (MEG) in humans to test the hypothesis that early novelty responses can be accelerated by reward. Critically, as in nonhuman primate studies of recognition memory, in experiment I the correct detection of either novel or familiar images was rewarded with £0.50, whereas in experiment II subjects discriminated novel from familiar images in the absence of reward.
In experiment I, experimental blocks in which novel images signaled monetary reward (CS+) and familiar images signaled no reward (CS−) alternated with blocks in which familiar images served as CS+ and novel images as CS−. Subjects were informed about the contingency before the beginning of each block and indicated via a button press with their right-hand index or middle finger whether they “prefer” or “do not prefer” the presented image based on the known contingency (A). Only correct “I prefer” responses following a CS+ led to a win of £0.50, whereas (incorrect) “I prefer” responses following CS− led to a loss of £0.10. Both correct “I do not prefer” responses following a CS− and (incorrect) “I do not prefer” responses following a CS+ led to neither win nor loss. Importantly, as in nonhuman primate studies, making correct preferences was only possible after correctly discriminating novel and familiar stimuli 
In experiment II, subjects indicated the novelty status of images also via a button press with either the index or middle finger of their right hand and the same response apparatus as in experiment I (C). In order to match the alternation of response contingencies across blocks in both experiments, the contingency between response finger and novelty status changed from block to block and was also announced at the beginning of each block (Experimental Procedures
; see also Figure S1
available online). Furthermore, in both experiments, response-related feedback was given not on a trial-by-trial basis but after the end of an experimental block, and subjects were instructed to respond as accurately as possible as soon as they could classify the stimuli. (See Experimental Procedures
for further details about both experiments.)
Behaviorally, subjects' memory performance was equally accurate in both experiments (discriminability index d′ > 2.1) without a significant response bias (β not different from 1) (). However, responses to novel as well as familiar images were faster in experiment I than in experiment II (p < 0.05 by two-sample t test; see legend). MEG data were averaged to event-related magnetic fields (ERFs) and statistically analyzed with SPM8 (Wellcome Trust Centre for Neuroimaging). The time course of ERF differences between conditions was assessed using a priori time windows of interest: 85–115 ms, 115–150 ms, 150–200 ms, 200–500 ms, and 500–700 ms. The first time window (85–115 ms) was motivated by aforementioned animal findings [6, 7]
; 150–200 ms, 200–500 ms, and 500–700 ms were chosen based on novelty effects reported in humans [2–5]
. In priming studies, but not in recognition memory studies, immediate stimulus repetition responses have also been reported in an early time window from ~100 to 150 ms [18, 19]
. Although these early priming effects were reliable at repetition intervals of less than 100 ms and were not novelty responses (because in these studies, each item was prefamiliarized), for completeness we also considered the time window of 115–150 ms. This was also done to fully characterize the temporal evolution of novelty effects.
Averaged ERFs for each condition per subject and time window were entered into a second-level random-effects analysis (i.e., experiment I, 2 × 2 analysis of variance [ANOVA] with the factors novelty [novel, familiar] and reward [rewarding, not rewarding]; experiment II, one-way ANOVA with the factor novelty [novel, familiar]). The earliest time window (85–115 ms) revealed a main effect of novelty over left temporal sensors for experiment I in the absence of any interaction (B), but no effects for experiment II (D). Instead, we observed a “classical” pattern in experiment II of a main effect of novelty for the time window 200–500 ms over frontal sensors (B). The same time window (200–500 ms) in experiment I revealed a main effect of reward over right frontal sensors (A); this is consistent with the fact that in experiment I, the preference judgment followed an initial old/new discrimination. For the late effect (500–700 ms), both studies revealed a main effect of novelty over either parietal (experiment I; C) or frontocentral sensors (experiment II; D). Furthermore, for the time window 115–150 ms, experiment I revealed a main effect of novelty over left temporal sensors (Figure S2
), but there was no such effect for experiment II. We did observe a main effect of novelty for experiment II for the time window 150–200 ms at a liberal threshold (p = 0.01, uncorrected; F = 9.07; Figure S3
), but there were no effects for this time window in experiment I. For a complete list of all effects, see Table S1
Statistical Parametric Maps of F-Statistics and ERFs
In a subsequent step, we directly compared the novelty responses in both experiments from two sensor locations (temporal and frontal) and time windows (85–115 and 200–500 ms) via 2 × 2 ANOVAs with the factors novelty (novel, familiar; averaged over reward status in experiment I and response finger in experiment II) and the between-subject factor experiment (experiment I, experiment II). In the early time window (85–115 ms; ERFs extracted from the peak over temporal sensors) (B), a significant interaction between novelty and experiment (F(1,26) = 5.29, p = 0.03; as well as a main effect of novelty, F(1,26) = 4.39, p < 0.05) with significant differences between ERFs for novel and familiar images in experiment I (p < 0.01 by t test) but not experiment II (p > 0.85 by t test; B and 1D) showed a dissociation between early novelty effects and experiments. At the later time window (200–500 ms), there was a main effect of novelty over frontal sensors (F(1,26) = 7.53, p < 0.05) but no significant interaction between novelty and experiment (p > 0.05), suggesting a tendency toward novelty effects in both experiments. The direct comparison of the novelty effects (difference ERFs) between the early time window in experiment I (85–115 ms) and the later time window in experiment II via two-sample t test in SPM did not show any significant differences (p = 0.005 with either unscaled or scaled values). This suggests no statistically significant topography differences between the earliest time window of both experiments. (See Supplemental Data
for analyses regarding the topography of old/new effects within experiments.)
Our findings show that in recognition memory tasks, complex novel and familiar stimuli can be neuronally discriminated in humans as early as 85 ms after stimulus onset if detecting either the novel or the familiar stimulus is rewarded. This is ~70 ms earlier than the earliest novelty effects reported in previous human recognition memory studies, which were in the range of ~150 ms [2–5]
. Importantly, the reward status of stimuli was also already signaled from 200 ms onward. This is remarkably rapid, because in order to signal whether a stimulus predicted reward or not in our paradigm, it was first necessary to determine whether a stimulus was novel or familiar. In fact, the ~100 ms time difference between the onset of the early novelty response and the reward response is compatible with the possibility that reward status was determined on the basis of the early novelty signal. Therefore, our data strongly suggest that even these very early novelty signals can contribute to the rapid retrieval of behaviorally relevant contingencies. Apart from this novelty-reward contingency, it remains to be established how these novelty signals contribute to conscious or unconscious 
forms of recognition memory decisions.
B and B suggest that the earliest ERF novelty responses in both experiments had different topographies (temporal in experiment I, frontal in experiment II). These responses also had different polarities (novel more positive versus familiar more positive), an interesting parallel to findings in nonhuman primates showing that monkey prefrontal novel-familiar response differences have a latency of ~200 ms and are reversed in polarity compared to early (~80 ms) temporal responses 
. However, statistical comparisons of experiments I and II show that topographic differences (~85 ms versus ~200 ms; across all contingencies in experiment I as well as experiment II) were not significant. Hence, it remains unclear whether reward motivation merely increased the speed of neural novelty processing or whether it had differential facilitatory effects on the generators of the temporal and frontal novelty responses.
Remarkably, the early novelty effect occurred despite delayed reward feedback and independently of whether novel or familiar items were rewarded (no interaction between novelty and reward), suggesting that facilitation in the context of reward cannot be eliminated by experimentally counterbalancing the contingencies between novelty and reward. It is also remarkable that the effect occurred in the absence of differences in discriminability (d′) and response bias (β).
Behaviorally, subjects responded faster to novel and familiar items in experiment I as compared to experiment II (). This finding fits well with recent observations of enhanced energization of action through reward as expressed in faster reaction times 
and increased response vigor [13, 15, 16]
. In contrast, discriminability (d′) between novel and familiar items did not differ between experiments. Hence, our data show that reward motivation at retrieval accelerates access to memory representations but does not substantially change the quality of representations accessed. Reward-related improvements in memory accuracy are more likely to be seen for an explicit reward manipulation at encoding (see for example [14, 21]
) rather than retrieval.
Taken together, our results show that reward motivation accelerates neural novelty processing and provide a framework for understanding differences between human and nonhuman primate studies of recognition memory. More generally, our findings indicate the importance of studying effects of motivation on the chronometry and functional anatomy of cognitive processes. Although the precise physiological mechanisms for reward-motivated facilitation of very early novelty processing remain to be established, one possibility is that elevated levels of dopamine in the context of reward may play a role 
. Given the very early onset of novelty responses, context-driven tonic effects of dopamine, for example related to behavioral or attentive set [22, 23]
, are likely to provide a more plausible mechanism than stimulus-driven phasic effects of dopamine [16, 24]