The preparatory processes or brain activity before an event do not only influence our perception, attention or motor responses but also influence higher-level cognitive
49
processes such as working memory (Bollinger et al., 2010), task-switching (Poljac &
Yeung, 2014) or problem-solving (Kounios et al., 2006). In the memory domain, Tulving (1983) proposed that an appropriate cognitive set supports to treat an event as a retrieval cue to retrieve stored episodic information. The development of neuroimaging techniques allow to study the neural correlates of this cognitive set. Neural activity preceding an event influences the efficiency with which information is retrieved from long-term memory storage (Rugg & Wilding, 2000; Wilding & Ranganath, 2012). A recent study found that pre-stimulus theta power increase leads to correct source memory retrieval (Addante, Watrous, Yonelinas, Ekstrom, & Ranganath, 2011). From the first study on brain activity before an event that predicts memory encoding (Otten et al., 2006), there has been a good number of studies on investigating how neural correlates before an event influence memory formation.
Using the subsequent memory paradigm, it has been demonstrated that the brain activity elicited by an event is important to memory encoding of that event. Otten et al. (2006) found that neural activity elicited by a cue before an event could also predict subsequent memory performance. In their study, ERP waveforms elicited by the pre-stimulus cue were categorised by subsequent memory performance. The waveforms before subsequently recognised words were significantly more negative than the waveforms before subsequently forgotten words over frontal scalp sites. This frontal negative-going pre-stimulus subsequent memory effect was only found after a cue that indicated a semantic encoding task needed to be done on the upcoming word. In a later fMRI study, a monetary incentive cue was presented before a target event to indicate reward value if the event was memorised (Adcock et al., 2006). Intervals between a reward cue and a target were jittered between 2.5 and 6.5 s to avoid BOLD signals overlap across trials. It was found that only after high reward cues, activation in hippocampal areas before target event onset predicted later memory performance. An
50
MEG study showed a pre-stimulus encoding-related theta power increase effect regardless of whether the encoding task was semantic or non-semantic (Guderian et al, 2009).
What is the functional role of encoding-related pre-stimulus activity? In the study by Otten et al. (2006), such encoding-related pre-stimulus activity was evident only when pre-stimulus cues indicated a semantic encoding task about upcoming words.
Therefore, they suggested the frontal negative-going pre-stimulus subsequent memory effect reflected a semantic preparatory process. Participants may have used the information of the cue to engage a more elaborated semantic preparation for later successful formation of memory representations. Further, this frontal negative-going pre-stimulus subsequent memory effect is not modality-specific (Otten et al., 2010).
Otten et al. (2010) used cues to indicate the modality of upcoming words. In a supplementary result figure from Otten et al. (2006) study, a visual cue signalling a presentation of an auditory word elicited larger P2 amplitude than a visual cue signalling a presentation of a visual word. Switching modalities between visual cues and auditory words requires more attention, which may impair preparation for encoding upcoming auditory words. Therefore, Otten et al. (2010) kept the modalities of cues and stimulus consistent. As a result, frontal negative-going pre-stimulus subsequent memory effects were found in both auditory and visual conditions. Such effects may reflect a wide role of semantic preparation in forming a verbal memory representation. This frontal negative-going pre-stimulus subsequent memory effect is also found when using free recall test (Galli et al., 2012; 2013). In both studies, the effects were evident only when there was opportunity to employ more elaborate strategies to memorise words, which suggests a role of the frontal negative-going pre-stimulus subsequent memory effect in semantic preparation.
51
Similar to encoding-related activity after an event, encoding-related activity before an event might be qualitatively different if different anticipatory processes are engaged. In an ERP study, Gruber and Otten (2010) used monetary reward cues to indicate reward value when a word was memorised. A more sustained positive-going widespread scalp distributed pre-stimulus subsequent memory effect was found only when the cue indicated a high reward value. This suggests that high monetary motivated preparation for verbal memory formation might be a different mechanism from semantic preparation. High monetary motivation also activated dopaminergic midbrain and MTL before stimulus onset and the activations predicted better subsequent memory performance (Adcock et al., 2006). In addition, high monetary motivation increased pre-stimulus theta power to modulate memory encoding (Gruber et al., 2013). Only when pre-stimulus cue indicated high reward value, frontal theta power increased before remembered words. Positive-going pre-stimulus subsequent memory effects were also found when anticipating an emotionally negative scene (Galli et al., 2011; 2014). Pre-stimulus cues indicated the valence of upcoming pictures. ERP waveforms elicited by a cue indicating that the upcoming picture would be an emotionally negative scene were significantly more positive-going before remembered pictures than forgotten pictures. It was suggested the preparatory processes engaged in encoding of a negative picture might be similar to the preparatory processes of memorising a high rewarded word as reward and emotion might share similar motivational mechanisms (Galli et al., 2011).
Galli et al. (2011) also suggested that the frontal negative-going pre-stimulus subsequent memory effect probably reflects different preparatory processes that are more cognitively oriented. In an ERP study, Padovani, Koenig, Brandeis and Perrig (2011) directly compared semantic preparation and emotional preparation. A pre-stimulus cue indicated whether participants needed to make semantic or emotional decisions about upcoming words. A frontal negative-going pre-stimulus subsequent
52
memory effect was found before semantic decisions while a more central positive-going pre-stimulus subsequent memory effect was found before emotional decisions. The results suggest there might be multiple preparatory neural mechanisms supporting memory formation.
The above findings indicate that encoding-related pre-stimulus activity might not reflect random fluctuations. Instead, such activity may reflect a beneficial state for later encoding, similar to the role of pre-stimulus activity in the perception domain (Busch, Dubois, & VanRullen, 2009; Hanslmayr et al., 2007; Mathewson et al., 2009).
In contrast to that domain, however, encoding-related pre-stimulus activity might be under strategic control (Galli et al., 2013; Gruber & Otten). Strategic control is often recruited in executive functions such as prospective memory, task switching or the monitoring of conflicts when cognitive resources are limited (Bartholow et al., 2005;
Botvinick, Braver, Barch, Carter, & Cohen; Braver, Reynolds, & Donaldson, 2003;
MacDonald, Cohen, Stenger, & Carter, 2000; McDaniel & Einstein, 2001). Strategic control should be self-initiated by an individual and effortful (MacDonald et al., 2000;
McDaniel & Einstein, 2001). In the domain of long-term memory, Kuhl and Wagner (2009) suggested that strategic control processes help resolve conflicts between memory representations and irrelevant representations. The control processes play a role in selecting what needs to be encoded, stored and retrieved (Benjamin, 2007). For example, strategic processes can be reflected in time allocation when the difficulty of to-be-encoded items is manipulated (Son & Metcalfe, 2000). More study time is allocated to more difficult items. In addition, the strategic selectivity in encoding can be reflected by different values of to-be-encoded items (Castel, Benjamin, Craik, & Watkins, 2002).
Participants tend to remember more words with higher point values to get higher sum scores. These findings suggest that participants strategically allocate limited cognitive resources to encode items to adjust performance.
53
In a similar vein, encoding-related pre-stimulus activity does not seem to reflect automatic processes. Such pre-stimulus activity is only found under certain circumstances, such as semantic preparation or preparation for encoding negative scenes.
This suggests that encoding-related pre-stimulus activity might reflect effortful processes. Encoding-related pre-stimulus activity is also sensitive to the availability of cognitive resources. The activity is engaged in helping memory encoding only when sufficient cognitive resources are available (Galli et al., 2013). When participants have to switch modality between pre-stimulus cues and to-be-encoded items, no significant pre-stimulus subsequent memory effect was found in encoding auditory items (Otten et al., 2006). When the modality was consistent between pre-stimulus cues and to-be-encoded items, pre-stimulus activity helped to encode auditory words (Otten et al., 2010). Finally, preparatory processes are self-initiated by an individual. Pre-stimulus activity is engaged when preparing for encoding a word in a high, but not low, monetary reward condition (Gruber & Otten, 2010). Therefore, not only processes after a stimulus are controlled strategically, but also the processes before a stimulus to aid successful encoding.
Some fMRI studies found that the critical brain regions for pre-stimulus subsequent memory effects depend on the information that pre-stimulus cues hold.
Uncapher et al. (2011) found that activation of the dorsal parietal cortex preceding stimulus predicts successful encoding by using valid spatial pre-stimulus cues. Pre-stimulus activity in the amygdala is also positively correlated with later recognition performance when a cue informs of an aversive upcoming picture (Mackiewicz et al., 2006). Most fMRI studies showed the MTL was more activated prior to remembered event (Adcock et al., 2006; Mackiewicz et al., 2006; Park & Rugg, 2010). An MEG study (Guderian et al., 2009) and an iEEG study that recorded directly from the hippocampus (Fell et al., 2011) both have observed increases in theta power before
54
word onset associated with successful memory encoding. Both of the two studies suggest that the MTL theta activity reflected contextual activation before event onset.
Larger theta power in the MTL could pre-activate information of episodic context for binding with the later item (Fell et al., 2011; Guderian et al., 2009).
Yoo et al. (2012) shows that the Parahippocampal Place Area (PPA) was less activated before remembered scenes than forgotten scenes. Then, they presented a scene after the PPA BOLD signals were larger or less than a pre-defined reference signal.
Memory performance for the scenes presented after less PPA activity was significantly better than for the scenes presented after larger PPA activity. This study suggests that memory performance can benefit if an appropriate neural context is in place. However, there have been very few studies describing how memory benefits from appropriate preparation. Hanslmayr, Leipold and Bäuml (2010) used a think / no-think paradigm and found that when pre-stimulus cues informed the think / no-think task in advance, the proportion of forgotten items following ‘no-think’ cue were significantly larger than forgotten items presented with ‘no- think’ cue spontaneously. Nevertheless, they did not find a better performance for remembering when there was pre-stimulus cue due to ceiling effect of the task. There are three studies showing better memory performance benefitted from informative category cues, valid spatial cues or a longer pre-stimulus interval (Bollinger et al., 2010; Summerfield & Mangels, 2006; Uncapher et al., 2011).
They all suggest this is due to more attention to upcoming events modulated by top-down control. Galli et al. (2013) found pre-stimulus subsequent memory effects only when sufficient cognitive resources were available before encoding a word. However, overall memory performance did not show any advantage when sufficient cognitive resources were available during preparation. These results suggest that overall memory performance is not necessarily improved when pre-stimulus activity is engaged before encoding. If encoding-related activity before an event is a determinant of successful
55
retrieval, then memory performance should also depend on interactions between encoding-related activity before and after an event, consolidation and retrieval.
In summary, so far at least two kinds of pre-stimulus ERP activity have been found to predict successful encoding. Frontal negative-going pre-stimulus subsequent memory effect is thought to reflect semantic preparation. Positive-going widespread pre-stimulus subsequent memory effect may suggest motivational mechanisms. The critical brain region for encoding-related pre-stimulus activity is suggested to be the hippocampus. Only a few studies show differences in overall memory performance between informative cues and non-informative cues or between with pre-stimulus cues and without stimulus cues, which suggests that presence of encoding-related pre-stimulus activity does not have to enhance later memory performance as other memory stages may compensate later overall memory performance (Galli et al., 2013).