NORTHWESTERN UNIVERSITY
Validating Dissociations between Human Memory Subtypes: Conceptual Priming, Familiarity, and Recollection
A DISSERTATION
SUBMITTED TO THE GRADUATE SCHOOL IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
for the degree
DOCTOR OF PHILOSOPHY
Field of Neuroscience (Interdepartmental Neuroscience Program)
By
Joel Lawrence Voss
EVANSTON, ILLINOIS
June 2007
2
ABSTRACT
Validating Dissociations between Human Memory Subtypes: Conceptual Priming, Familiarity, and Recollection
Joel Lawrence Voss
A comprehensive understanding of human memory requires both cognitive and neural descriptions of memory processes along with a conception of how memory processing drives behavioral responses and subjective experiences. Noninvasive neuroimaging techniques have greatly extended our understanding of the functional characteristics of human memory, and how neural events give rise to different memory subtypes. Nonetheless, a great deal of uncertainty has clouded distinctions between hypothesized explicit memory processes termed recollection and familiarity, and a hypothesized implicit memory process termed conceptual priming. In this thesis, I specify this problem via a thorough review of findings from one form of neuroimaging, the recording of event-related brain potentials, and then outline a theoretical stance to justify why delineating these forms of memory has been particularly difficult from a neural perspective. I then present data from four neuroimaging experiments that clarify the neural relationships between these three expressions of memory. These neuroimaging experiments all employ novel behavioral paradigms in which multiple behavioral measures of memory subtypes substantiate the patterns of neuroimaging data, thus allowing valid conclusions to be made regarding the correspondence between behaviorally-indicated memory subtypes and neural measures. Taken together, results from these experiments indicate (1) that conceptual priming and familiarity are distinct forms of memory that are unlikely to result from the same neural processing, and (2) that familiarity and recollection, although phenomenologically distinct, are not qualitatively different in their neural implementation as revealed by electrophysiology, and thus likely reflect varying levels of explicit memory rather than distinct retrieval processes.
3 Contents
Part One: Introduction 6 1. Human Memory Subtypes 7 2. ERP Recording and Analysis 8 3. Quantifying ERPs 11 4. Advantages and Disadvantages of the ERP Technique 13
Part Two: A Review of ERP Findings 5. ERPs and Memory Encoding 16 6. The Dm Approach 16 7. Intracranial Dm Findings 21 8. ERPs and Memory Retrieval 22 9. ERP Correlates of Recollection and Source Memory 24 10. ERP Correlates of Post-Retrieval Processing 27 11. Difficulties Identifying ERP Correlates of “Pure” Familiarity 29 12. Using ERPs to Contrast Memory Subtypes 32 13. Perceptual Priming and Recognition Memory 33 14. Conceptual Priming 36
Part Three: Existing Evidence against a Unique ERP Signature of Familiarity 15. Familiarity and Conceptual Priming 38 16. Variables That Cannot Differentiate Familiarity from Conceptual Priming 43 17. Variables That Can Potentially Differentiate Familiarity from Conceptual Priming 50 18. Direct Tests That Have Failed to Validate FN400 Correlates of Familiarity 53
Part Four: ERP Correlates of Familiarity and Conceptual Priming for Famous Faces 19. Rationale 57 20. Methods 58 21. Results 63 22. Discussion 70
Part Five: fMRI Correlates of Familiarity and Conceptual Priming for Famous Faces 23. Rationale 74 24. Methods 74 25. Results 78 26. Discussion 81
Part Six: ERP Correlates of Familiarity and Conceptual Priming for Minimalist Visual Shapes 27. Rationale 85 28. Methods 85 29. Results 90 30. Discussion 98 4
Part Seven: ERP Correlates of Familiarity and Conceptual Priming for Words 31. Rationale 102 32. Methods 102 33. Results 105 34. Discussion 111
Part Eight: Concluding Remarks 115
Bibliography 124
Illustrations
Tables
1. Memory Taxonomy 140 2. Summary of fMRI Activation Clusters for Famous Faces 141 3. Conceptual Priming for Squiggles 142 4. Recognition Performance for Uncommon Words 143 5. Conceptual Priming for Uncommon Words 144 6. ERP Formal Comparisons for Uncommon Word Recognition 145
Figures
1. Event-Related Brain Potentials 146 2. The Subsequent Memory Paradigm 148 3. Representative LPC Encoding Effects 149 4. Representative LPC Retrieval Effects 151 5. Representative Late Frontal Effects 152 6. Contrasting Recollection and Perceptual Priming 153 7. Paradigm Used to Examine ERPs to Famous Faces 154 8. ERP Correlates of Memory for Famous Faces 156 9. ERP Correlates of Familiarity for Famous Faces 158 10. Paradigm Used to Examine fMRI Responses to Famous Faces 159 11. fMRI Correlates of Memory for Famous Faces 160 12. fMRI Impulse Response Functions 162 13. Squiggle Stimuli 163 14. Recognition Performance for Squiggles 164 15. ERP Correlates of Recognition for Squiggles 165 16. ERP Correlates of Recollection and Familiarity for Squiggles 166 5 17. ERP Correlates of Conceptual Priming for Squiggles 167 18. Correlations between ERP and Behavioral Correlates of Recognition 168 19. Electrode Locations 169 20. ERP Correlates of Conceptual Priming for Uncommon Words 170 21. ERP Correlates of Recognition for Uncommon Words 171
Appendices
1. Biographical Cues 172 2. Uncommon Words 175 6 Part One: Introduction
This thesis describes the use of noninvasive neuroimaging measures to probe the neural
substrates of human memory. After outlining a taxonomy that has proven useful in characterizing
various memory subtypes, I review the literature on event-related brain potential recordings and
the insights they have provided with regard to substantiating distinctions within the taxonomy. I
then highlight several border areas within this taxonomy that need further clarification: those between hypothesized memory processes known as recollection, familiarity, and conceptual priming. Comparisons between these forms of memory constitute the primary theoretical focus of this thesis, as described in Section 15. A literature review examining the difficulties associated with delineating these forms of memory follows. Finally, I present data from four neuroimaging experiments that make progress towards demarcating these boundaries, and suggest how further
insights can be gained in this area.
A portion of the sections herein derive from work that has been published previously, is
currently under peer review, or is in preparation. The general review of electrophysiological
studies of human memory presented in Part 1 and Part 2 contains information presented in (Voss
& Paller, in press). Information presented in Part 3 echoes an argument presented in (Paller,
Voss, & Boehm, in press). Part 4 describes an experiment presented in (Voss & Paller, 2006).
Part 5 describes an experiment presented in (Voss, Reber, Mesulam, Parrish, & Paller, under
review). Part 6 describes an experiment presented in (Voss & Paller, 2007). Part 7 describes an
experiment presented in (Voss, Lucas, & Paller, in preparation). In presenting these findings here
as a coherent whole, it is possible to identify theoretical links that otherwise remain elusive, and
thus to clearly identify these experiments’ contributions to the understanding of human memory
at-large. 7 1. Human Memory Subtypes
Remembering is a multifaceted cognitive activity that can be partitioned into a set of component processes. These processes are sometimes interrelated and sometimes operate independently, and are expressed in different combinations under different circumstances. These distinct memory functions have been uncovered largely by studying the memory capabilities of both healthy individuals and patients with selective memory impairments, and can be assessed using different specialized memory tests. A variety of theoretical schemes have been used to categorize the memory phenomena measured in these tests, emphasizing either behavioral, cognitive/representational, neural, or subjective criteria (White, in press). Taxonomies of memory have guided research into fundamental questions about memory. However, a comprehensive understanding of memory must go beyond taxonomy by defining each component process in both cognitive and neural terms, by specifying the relationships between cognitive and neural descriptions, and by showing how neurocognitive processes produce memory behavior and associated conscious experiences.
Amnesic patients have selective impairments in declarative memory, the ability to remember facts and events from the past, as assessed in recall and recognition tests. In contrast, other categories of memory phenomena, as listed in Table 1, are not impaired in amnesia (Grill-
Spector, in press; Schacter, in press; Squire, in press). Expressions of declarative memory tend to coincide with the potential for making the metamemory judgment that memory is being expressed—the awareness of remembering. For these reasons, declarative memory is usually regarded as fundamentally distinct from other expressions of memory.
Information can also be held in awareness for an extended period of time, while rehearsed and/or manipulated. Nonetheless, the emphasis here is on long-term memory 8 phenomena that take place when information that was initially encoded is later brought back to mind after a delay, which is what William James (1890, pp648) termed secondary memory. [For a current summary of research on primary memory or working memory, see D’Esposito (in press)]
The neural substrates of remembering can be examined in healthy human volunteers using a variety of noninvasive neuroimaging techniques. In particular, recordings of event- related potentials (ERPs) have been used to monitor the activity of the brain during memory tasks. These experiments have made significant headway in identifying neurocognitive processes that are responsible for memory and in specifying processes engaged in association with different memory feats.
The goal of this introduction is to examine how ERP research has shed light on various processes. Memory processes that contribute to declarative memory are emphasized, but the related memory phenomena of priming is also included (Section 13). Indeed, the argument presented in subsequent chapters focuses on the border between declarative memory and priming.
2. ERP Recording and Analysis
Neurons in the human brain generate electric fields that vary moment to moment. When these electric fields are sampled via recording electrodes connected to an amplifier system, the resultant electroencephalographic record—the EEG— shows voltage changes over time and can provide indications of the functioning of networks of neurons as cognitive processing unfolds.
EEG recordings from electrodes placed on the scalp are used in a variety of clinical and research contexts. 9 To examine EEG activity associated with stimulus processing, signal-averaging methods
applied to EEG recordings can be used to produce ERPs. Whereas ongoing EEG signals can vary
in magnitude on the order of tens of microvolts over a few seconds, ERPs can be extracted so as
to observe signals much smaller than one microvolt. When a set of EEG responses that are time-
locked to a particular class of stimuli are averaged in this way, EEG signals unrelated to stimulus
processing tend to decrease because they do not occur at a consistent time relative to the time of
stimulus onset. ERPs thus reveal electrical signals produced during the course of stimulus
processing, to the extent that such signals are not averaged out due to temporal variability. Signal
averaging can be performed with respect to any class of repeated event occurring at a known time. ERPs elicited by stimulus events are viewed as a time series of voltage changes following stimulus onset, which is conventionally referred to as “time 0.” ERP waveforms thus consist of a series of positive and negative deflections, the timing and waveshape of which vary with the nature of the stimuli and the neural operations performed in response to the stimuli. Some ERPs, such as the brainstem auditory evoked potentials produced during the first 10 ms after a click, have very small amplitudes and require hundreds of stimulus repetitions to obtain a signal-to- noise ratio sufficiently high to permit reliable quantification. Other ERPs, such as some associated with cognitive processing, can be several microvolts in amplitude and occur over a time interval of several hundreds of milliseconds. Typically, averaging over 30-100 stimulus events is required to observe reliable effects of experimental variables in psychological experiments, depending on the amplitude and reliability of the EEG signals in question and on the presence of other EEG signals and EEG artifacts of various sorts (Luck, 2005; for further methodological details, see M. D. Rugg & Coles, 1995). 10 ERPs can be most readily measured by examining the positive and negative deflections
(peaks and troughs) that occur at various poststimulus time points. However, the entire
waveform is most likely composed of the summation of neural activity from many distinct sets
of neurons in the service of many different functions. Ideally, a complex ERP waveform would
be decomposed into a series of ERP components, each of which bearing a unique and systematic
relationship to a unitary neurocognitive function. In practice, however, the component structure
of an ERP waveform is difficult to discern. Given this strict definition that requires a component
to be identified with a unique neurocognitive function, it would be unwise to accept the
assumption that each deflection corresponds to a particular component. There may be some cases
when a hypothetical component may be adequately measured by examining a deflection. In other
cases, deflections are based on the summation of multiple components, which themselves are
unknown, such that the amplitude and latency of the composite peak does not provide a valid
characterization of any of the individual components. Accordingly, the identification and
measurement of specific ERP components can be problematic.
ERP waveforms can nonetheless be quantified in several different ways. Putative components can be identified based on a combination of factors, including latency, polarity, amplitude, distribution, and most importantly, relationships to experimental parameters. ERPs can also be quantified for specific latency intervals without making a priori assumptions about
the components that might be present, but with an emphasis instead on differences between
experimental conditions. When this approach is followed, the experimental manipulations play a critical role in focusing the analysis on neurocognitive functions that can be manipulated across
conditions. Valid conclusions can be drawn based on such analyses, given that some conclusions are orthogonal to the challenge of determining whether specific aspects of the ERP display a 11 convincing correspondence with ERP components that have been described previously.
Although component identification can be informative, it may not be feasible in memory
paradigms when a large number of components overlap with each other in the same time range.
Indeed, when subjects engage a wide variety of cognitive transactions over an extended time
interval, as is likely the case in many of the interesting paradigms cognitive neuroscientists
choose to study, it can be misleading to assume that only a very small number of components
have been produced. When a large number of ERP components occur simultaneously, specific
ERP components cannot always be isolated from one another and separately characterized.
Accordingly, many ERP investigations in cognitive neuroscience no longer exemplify the strict component-centered approach, wherein a chief experimental goal was to understand an
ERP component per se. Instead, difference-centered approaches have become prevalent, whereby experimental variables are manipulated based on theory-driven goals concerning specific neurocognitive functions. This shift from a focus on known ERP components to a focus on the thorough understanding of relationships between cognitive operations and neural events has facilitated a greater dialogue between ERP experimenters and those working with different methodologies. In the context of memory research, bringing together a variety of methods in cognitive neuroscience has been responsible for significant progress.
3. Quantifying ERPs
In ERP investigations of human memory functions, a central analysis question in any experiment is often to determine whether two ERPs differ reliably from one another. In other words, the experimenter may ask whether two or more hypothetical psychological processes are associated with reliably different electrical signals. Experimental contrasts are often based on 12 comparing two conditions distinguished by a task manipulation, stimulus factors, response
factors, or some combination. A difference between ERPs can then be described and displayed
(Figure 1).
An ERP difference may be statistically significant over a certain time interval. At any
given latency, the difference can have a positive or negative polarity. Across a recording epoch,
amplitude will vary with a particular wave shape. When an ERP difference between conditions is
characterized in this manner, it may appear to correspond to a systematic enhancement of one
ERP deflection over a discrete time interval, or it may appear to encompass a different time interval with a unique wave shape.
Another important facet of an ERP difference is the distribution of the potential field across multiple electrode locations on the scalp. This topographic information can help investigators make inferences regarding the responsible neural generators. Such inferences involve many assumptions. Models of electric currents and volume conduction of the head can be used to estimate the scalp topography that would be produced by activity at a certain location in the brain. Despite straightforward procedures for solving this so-called forward problem, the inverse problem of determining the brain sources based only on the scalp topography is not soluble, because many different configurations of intracranial generators can produce the same field on the scalp. Various ERP source-modeling procedures can nevertheless be used when considering the anatomical location of the sources of ERPs, although drawbacks of inferring brain sources based on scalp recordings have been heavily debated (Kutas & Dale, 1997;
McCarthy & Wood, 1985; Urbach & Kutas, 2002; Wilding, 2006). When theorizing about ERP sources in the brain, it is therefore extremely helpful to bring multiple sources of evidence to bear on understanding the relevant brain structures or systems. 13 4. Advantages and Disadvantages of the ERP technique
ERP methods are noninvasive and relatively inexpensive compared to other neuroimaging
methods. As just described, some limited information about relevant neural sources can be extracted from ERP distributions on the scalp. In most circumstances, other methods are preferable for precise neuroanatomical information. ERP waveforms comprise a time-series of voltages between a scalp electrode location and a reference electrode location, so both locations are relevant for determining ERP characteristics. Results from recordings using a mastoid or average-mastoid reference are emphasized here, because this recording method is used in a majority of the relevant memory studies. Like any reference that can be used during recording or
digitally created afterwards, the mastoid reference is not inactive, as potentials generated near
this location influence observed ERP effects in accordance with how currents are conducted
through tissue to electrode locations.
Like other neuroimaging methods, ERPs do not provide a full view of neural activity.
Rather, the EEG is sensitive to activity produced by a restricted set of neurons. These neurons
are ones activated synchronously such that extracellular fields produced by their activity can
summate. These fields must thus be generated by sets of neurons that are oriented together in
such a way to produce electrical fields that will conduct to distant locations where recording
electrodes are placed. Much neural activity may be electrically silent at the scalp, in the sense
that the EEG may not include signals from some of the active neurons engaged in relevant
processing.
A key advantage of ERP methods is that they provide measures of neural activity with
very high temporal resolution. The superior temporal resolution of ERPs makes them well suited
to examinations of neural events responsible for human memory, which can potentially be 14 monitored by ERPs on a millisecond-by-millisecond basis. Critical memory processes often unfold within the first second after exposure to a stimulus, and ERPs can allow these processes to be resolved in real-time and with randomized trial orders. Delivering trials in a predictable manner or blocking experimental trials, as required in positron emission tomography (PET) studies, can severely limit the range of memory phenomena amenable to examination. Extended intertrial intervals can also be undesirable to the extent that such procedures do not adequately constrain the timing of relevant cognitive events. In general, when cognitive events can be tightly time-locked to stimulus presentation and temporal blurring across trials decreased, ERP findings have better signal-to-noise ratios and arguably are most useful. Although randomized event- related designs with short intertrial intervals (Burock, Buckner, Woldorff, Rosen, & Dale, 1998) are feasible with functional magnetic resonance imaging (fMRI), ERP signals may be better for isolating brief neurocognitive events or, especially, a series of neurocognitive events that occur closely together in time.
Although ERPs provide temporal resolution that is unsurpassed by that of any other technique in cognitive neuroscience, other neuroimaging modalities that also provide high temporal resolution, including magnetoencephalography (MEG) and monitoring near-infrared optical signals. Such methods hold promise for substantial contributions to the memory literature in the future, and may also provide further neuroanatomical insights. Additional information can also be extracted from these signals or from EEG signals by conducting analyses in the frequency domain. Stimulus events produce reliable EEG oscillations that may reveal insights into neural activity that are complementary to those available from analyses in the time domain.
Oscillations that occur in phase with stimulus onset are generally apparent in EEG records and are thus called evoked rhythms. Other oscillations can be time-locked to an event but be out of 15 phase from trial to trial; these are called induced rhythms. A small but steadily increasing body
of literature concerns memory-related cyclic EEG activity (e.g. Duzel et al., 2003; Klimesch et al., 2001; Klimesch et al., 2006). Prospects for combining ERP methods with frequency-domain analyses of EEG activity thus hold great promise.
16 Part Two: A Review of ERP Findings
5. ERPS and Memory Encoding
Experiments examining long-term memory generally employ an encoding phase, during which subjects attempt to commit items to memory, followed after some delay by a test phase, during which the success of memory storage and retrieval is evaluated. ERP measures can be collected both at encoding and at test, informing accounts of the relevant neural processing required during these stages. Given that declarative memories change over time, it will ultimately be important to examine relevant processing that can take place at various times between initial encoding and later retrieval. Less work has been devoted to this challenge.
The focus of most ERP studies of memory has been on explicit memory for episodes.
Some studies have examined autobiographical memories formed outside the laboratory, or memories for general semantic knowledge learning over many years. The most commonly used paradigms concern memory for artificial events in a laboratory setting, such as viewing specific words or images. These studies are advantageous because the circumstances of acquisition can be carefully monitored and controlled. As laboratory studies move closer to accurately simulating real-life memory experiences, it is possible that the artificial nature of this research will become less problematic in placing limitations on interpretations.
6. The Dm Approach
One way to isolate brain events responsible for successful encoding is to acquire ERP responses during encoding and sort trials based on subsequent memory performance (Figure 2).
This general method was first reported with ERPs by Sandquist and colleagues (1980) and can 17 be traced back to earlier work using skin-conductance methods. Indeed, many sorts of neural signals can be used in subsequent-memory analyses. The term Dm has been used to refer to neurophysiological Difference measures found by sorting trials on the basis of subsequent
Memory performance (Paller, Kutas, & Mayes, 1987). This term provides a convenient way to refer to the various phenomena that can be demonstrated using these methods (e.g., Dm for free recall, Dm for recognition, Dm for pure familiarity), and also avoids prejudging whether the differences reflect variations in a known ERP component. Subsequent-memory analyses can be conducted with different encoding requirements, different types of memory tests, and different retention intervals — and all of these parameters may influence the results. Dm potentials observed with different task, stimulus, and subject parameters can presumably index various neurocognitive operations to the extent that these operations partially determine what information will later be remembered.
Observing reliable Dm effects generally requires that multiple criteria are met, including some inter-item variability in encoding strength such that a sufficient number of items are subsequently remembered and subsequently forgotten. Logically, Dm effects will be greater when ERPs index a larger difference between mean responses in these two conditions. Dm analyses can thus gain power when confidence or other graded measures of retrieval success are
considered. Ideally, there will exist a substantial polarization in successful versus non-successful
encoding operations. Furthermore, temporal dynamics must also be suitable, such that processing
that influences later memory performance is well time-locked to stimuli presented for encoding.
In many Dm studies, positive potentials maximal over parietal regions were found to be
greater for later-remembered items than for later-forgotten items at approximately 400-800 ms.
These effects have sometimes been attributed to ERP components such as P300 or the late 18 positive complex (Fabiani, Karis, & Donchin, 1986; e.g. Karis, Fabiani, & Donchin, 1984). In the past two decades, Dm has been observed in many paradigms, including tests of recall and recognition (Paller & Wagner, 2002; reviewed in Wagner, Koutstaal, & Schacter, 1999). A favored cognitive hypothesis about these Dm findings is that they reflect superior elaborative encoding for items later remembered. Semantic elaboration is well known to be effective for producing strong episodic memories, and deeper semantic elaboration often corresponds with larger late posterior potentials. The nature of Dm effects due to elaboration can differ depending on the nature of the elaborative processes, however, in that frontal slow waves rather than late parietal-maximum positive potentials have been observed when subjects attempt to remember lists of unrelated words by generating novel associations to create relationships among items
(Fabiani, Karis, & Donchin, 1990; Friedman, 1990). Creating novel associations likely requires a larger contribution from working memory processes supported by frontal cortex compared to elaborating on the inherent meaning of a single word.
In a recent study with faces, results revealed distinct Dm effects depending on the type of memory retrieval possible during the subsequent memory test (Yovel & Paller, 2004). As shown in Figure 3, a robust Dm that was bilaterally symmetric at posterior scalp locations was found to predict later recollection (when subjects could recall information previously associated with the face), whereas a smaller right-sided Dm predicted later familiarity (when subjects recognized the face but could retrieve no additional information).
Based on another type of Dm demonstrated in a few studies with words, it appears that left-frontal potentials starting at approximately 500 ms can be sensitive to the amount of information bound into a memory trace at encoding (reviewed in Friedman & Johnson, 2000).
The magnitude of these potentials may be proportional to the amount of information remembered 19 subsequently. One hypothesis is that these sustained positive potentials reflect the operation of
strategic encoding processes.
In addition to stimulus-evoked neural processing that leads to later remembering, tonic
brain activity that precedes the onset of a stimulus has also been linked to the efficacy of
memory encoding. For example, in a study reported by Otten and colleagues (2006), ERPs over
the front of the head were more negative for subsequently remembered versus subsequently
forgotten items in the interval preceding stimulus onset by several hundred milliseconds. These
electrophysiological findings have an important implication for investigations of encoding using
other methods such as fMRI; anticipatory as well as stimulus-locked neural activity could be
blurred together due to the poor temporal resolution of fMRI. Procedures are thus needed to
separate fMRI correlates of these dissociable cognitive events. Although anticipatory Dm
activity has not been extensively investigated, one compelling hypothesis is that it reflects
working memory control operations, given that such operations can lead to better encoding of a
declarative memory.
Gonsalves and Paller (2000b) showed that ERPs at encoding can reflect mistaken
memories as well as accurate ones. ERPs were found to reflect the vivid visual imagery when
subjects visualized the referent of a word, and this activity was found to be relevant to false
remembering. Trials were categorized according to whether or not subjects subsequently claimed
(erroneously) to have seen a picture of the visualized object, revealing that these posterior ERPs
were predictive of whether or not a false memory for that item would happen in the test phase.
ERPs thus revealed encoding activity that was partially responsible for the later mistake. In
contrast, large and more widespread ERPs were predictive of accurate memory for viewed
pictures. 20 On the whole, Dm effects may index a group of encoding operations that lead to superior memory, including detailed perceptual analysis, rote rehearsal, semantic elaboration, mental imagery, and so on. Further research is needed to clarify how Dm varies as a function of stimulus materials and type of memory test, and to precisely specify the relationship between these electrical measures and specific encoding operations. A fruitful approach for future neuroimaging studies, for example, would be to directly manipulate hypothetical mnemonic processes to determine the relative contribution of different operations to Dm effects. The usefulness of this approach for advancing explanations of memory formation can be exemplified by a recent fMRI investigation (Reber et al., 2002). Subjects were instructed to intentionally remember some words and intentionally forget others. By virtue of this directed-forgetting manipulation, fMRI correlates of the differential intention to remember were identified in addition to standard Dm effects based on retrieval success. fMRI Dm effects commonly include increased activity in both inferior prefrontal cortex and the medial temporal lobe. Although encoding condition and the probability of successful subsequent memory retrieval were correlated in this study, prefrontal activity was preferentially associated with encoding effort, whereas medial temporal lobe activity was preferentially associated with success. Results thus identified the specific contribution to encoding of mnemonic operations guided by the intent to remember a word and supported by left inferior prefrontal cortex. In this way, future studies could examine the gamut of effective encoding operations in order to use Dm analyses to dissect the neurocognitive processes that support memory formation.
21 7. Intracranial Dm Findings
Electrodes implanted into the brains of patients who are candidates for surgery to remove
an epileptic focus provide a special opportunity to combine the real-time temporal resolution of
the ERP technique with superior spatial localization. This approach has limitations, however, in
that activity can only be sampled from a limited number of brain regions, as electrodes are
placed only where required for clinical purposes. In addition, generalizability can be questioned
because recordings are made from a small number of individuals who have typically taken anti-
seizure medication for many years to try to control abnormal electrical activity in the brain.
Results can nonetheless be used to inform theoretical accounts of the neural basis of memory
formation.
In an ERP study reported by Fernandez and colleagues (2002), recordings were made
from two areas within the MTL, the hippocampus and an adjacent cortical region near the rhinal
sulcus. Based on research with amnesic patients and with nonhuman animals, the hippocampus together with adjacent parahippocampal, entorhinal, and perirhinal cortical regions have been considered as essential circuitry that is critical for the formation of declarative memories, in
conjunction with widespread neocortical regions that are ultimately responsible for memory
storage (Squire, 2004). Potentials recorded from the rhinal region (entorhinal and/or perirhinal
cortex) and hippocampus predicted subsequent free recall of visual words. The rhinal Dm peaked
approximately 400-500 ms after word onset and was thought to be associated with the extent to
which words were processed semantically. The hippocampal Dm, in contrast, started later and
was taken to reflect the successful binding of multiple features into memory following semantic
analysis. 22 These results support the notion that distinct MTL regions perform unique mnemonic operations. Other intracranial ERP studies have demonstrated a range of phenomena that may also reflect important memory functions (Allison, Puce, Spencer, & McCarthy, 1999; Engel,
Moll, Fried, & Ojemann, 2005; Grunwald, Lehnertz, Heinze, Helmstaedter, & Elger, 1998;
Guillem, Rougier, & Claverie, 1999; Heit, Smith, & Halgren, 1988; Paller & McCarthy, 2002;
Trautner et al., 2004; Viskontas, Knowlton, Steinmetz, & Fried, 2006). For example, a complex pattern of rhinal-hippocampal synchronization and de-synchronization was found via frequency- domain analyses of the same intracranial data (Fell et al., 2001). EEG synchronization in various other brain locations has also been observed to correlate with subsequent memory performance
(Sederberg, Kahana, Howard, Donner, & Madsen, 2003; Sederberg et al., 2006). Furthermore, single-unit firing patterns in the hippocampus have been shown to vary as a function of subsequent memory performance (Cameron, Yashar, Wilson, & Fried, 2001). Further studies are required to replicate and extend these various findings in order to elaborate on the information processing steps that are performed by neurons in each medial temporal lobe region, as well as to explore the temporal dynamics of interactions across multiple brain regions. ERP measures of memory formation, in combination with findings from these other methods, will be very important for delineating the distinct contributions to memory formation dependent on different brain processes and regions.
8. ERPs and Memory Retrieval
Whereas veridical memory performance is only possible if some information was initially encoded, the nature of memory is also a function of events taking place at the time of retrieval.
Furthermore, the most interesting distinctions between types of memory (e.g., declarative 23 memory and priming) and between memory processes (e.g., recollection and familiarity) are largely realized at retrieval. This is the time when one can engage in the conscious experience of remembering that can approximate re-living a past event.
In most ERP studies of memory retrieval, electrophysiological responses are recorded while recognition is tested. In a recognition test for episodes studied during an encoding phase, subjects must discriminate old (repeated) items from novel items. ERP correlates of episodic memory (sometimes termed episodic memory effects or old/new effects) are commonly identified by contrasting ERP responses elicited by old items to those elicited by new items.
Much effort in ERP studies of memory has been devoted to attempting to elucidate the specific memory processes that give rise to old/new effects.
Measures of neural activity obtained when people remember episodes are often interpreted in light of theories that posit two distinct recognition processes: recollection and familiarity (Yonelinas, 2002). Recollection involves the recognition that an event has occurred in the past along with the retrieval of specifics regarding the prior occurrence, thereby guiding the conscious experience of remembering. In contrast, familiarity denotes recognition of prior occurrence that remains unsubstantiated by retrieval of any specific detail. Familiarity can lead to a feeling of knowing in the absence of the ability to bring to mind any additional information.
Recollection and familiarity are connected with the concepts of source memory and item memory, in that source memory concerns the spatiotemporal context and various other features that can support recollection, whereas familiarity for a stimulus might be driven by retrieval limited to item memory.
An experimental procedure known as the remember/know paradigm has been employed extensively in attempts to identify neural correlates of recollection and familiarity. Subjects are 24 cued to introspectively classify their recognition of old stimuli as “remember” if specific study- phase detail is simultaneously brought to mind or as “know” if no such detail is retrieved. The
remember/know response categories have been taken as generic indices of recollection and
familiarity, respectively. Under some experimental conditions, however, remember/know
responses may correspond to varying degrees of memory strength instead of qualitatively
different memory processes (Eldridge, Sarfatti, & Knowlton, 2002). In addition, it is possible
that results obtained from this procedure are highly influenced by non-mnemonic variables, such
as the capacity to introspect accurately. Great care is thus needed in applying this method.
Confidence in results can be increased with convergent support from multiple methods, such as
with memory judgments based on source information.
Indeed, a useful approach to separating ERP components related to recollection and
familiarity is to test memory for specific source information based on the experimental context at
encoding. For example, subjects may encode words spoken by either a male or female voice, and
later be tested with visual words and asked to recall the original gender. In general, correct
source retrieval can be used to indicate recollection. Incorrect source retrieval with correct
recognition, however, is not always a good indicator of familiarity, given that recollection may
be supported by retrieval of information other than the specific source information in question.
Distinct ERP effects have been linked to three types of mnemonic processes associated
with memory retrieval. These findings are outlined in the following three sections.
9. ERP Correlates of Recollection and Source Memory
The most consistently reported finding in recognition studies is that ERP amplitudes to
old items are greater than those to new items from approximately 400-800 ms over much of the 25 scalp. These effects typically show a maximal difference over midline or left parietal scalp locations. The amplitude of these differences generally increases with increasing memory strength based on various behavioral indices (e.g., recognition confidence). Compared to
recognized items (hits), both old items that are forgotten (misses) and new items that are
correctly identified (correct rejections) tend to elicit smaller ERPs.
Early investigations of these old/new ERP effects endorsed a variety of hypotheses
concerning their functional significance, including associations with memory strength (R.
Johnson, Jr., Pfefferbaum, & Kopell, 1985), relative familiarity (M. D. Rugg, 1990), contextual
retrieval (Smith & Halgren, 1989), and processes that do not contribute to recognition judgments
(M. D. Rugg & Nagy, 1989). Despite this lack of consensus about the meaning of these ERPs
during these years, a common assumption was that the effects included modulation of two ERP components: N400 potentials and P300 potentials (e.g. Halgren & Smith, 1987).
An early study that convincingly associated ERPs with recollection utilized a levels-of- processing manipulation at study (Paller & Kutas, 1992). Behavioral results showed that this manipulation influenced recall and recognition performance, with superior memory following semantic encoding that required visual imagery than following encoding that focused attention on letter information. In contrast, the same level of priming (Table 1) was observed on an implicit memory test of word identification for both encoding tasks. ERPs recorded during the implicit memory test were compared between the two conditions defined by the task assigned at encoding, and corresponding differences were interpreted as ERP correlates of recollection. This
ERP difference based on encoding task began at a latency of 500 ms and was only present for words that were successfully identified in the test phase. Unlike typical old/new ERP effects, this effect could not be attributed to differences associated with priming because priming was 26 matched between the encoding conditions. Furthermore, behavioral evidence obtained at debriefing showed that subjects noticed that words from the encoding phase appeared during the word-identification test, even though this was irrelevant to their task. In other words, subjects were cognizant of specific contextual information with respect to some of the words in the word- identification test that repeated from earlier in the experiment. The authors thus inferred that incidental recollection took place during the test phase, particularly when word meaning had been encoded deeply, and that ERPs were sensitive to the differential processing associated with recollection.
Further studies using the same strategy in experimental design have substantiated the association between ERPs and recollection and extended the results to the use of other encoding tasks and memory tests (Paller, Kutas, & McIsaac, 1995), words presented in the auditory modality (Gonsalves & Paller, 2000a), and other types of stimuli such as faces (Paller, Bozic,
Ranganath, Grabowecky, & Yamada, 1999). These late parietal ERPs can thus be taken as signals of the successful retrieval of episodic memories linked with conscious remembering
(reviewed in Friedman & Johnson, 2000; Paller, 2000).
Results from remember/know as well as source-memory paradigms have also been used to associate late parietal ERPs with recollection. Late parietal ERP amplitudes are often found to be greater for remember compared to know responses and know responses compared to new trials. Importantly, there have not been convincing demonstrations that the distribution of late parietal ERPs differ between remember and know responses, indicating that these ERPs may index a neurocognitive operation that differs only quantitatively between recollection and familiarity conditions (e.g. Smith, 1993). Similarly, correct source judgments elicit greater late parietal amplitudes compared to incorrect source judgments, and incorrect source judgments 27 greater than new trials, without qualitative differences in scalp distribution (e.g. Trott, Friedman,
Ritter, Fabiani, & Snodgrass, 1999; Wilding & Rugg, 1996).
In addition to corroborating the connection between recollection and late posterior potentials, many experiments have targeted ERP correlates of source and item memory in order to understand contextual memory in its own right. Experimental contexts have included: speaker gender (Senkfor & Van Petten, 1998), performed, watched, or imagined actions (Senkfor, Van
Petten, & Kutas, 2002), background figures (Guo, Duan, Li, & Paller, 2006), spatial location
(Van Petten, Senkfor, & Newberg, 2000), and stimulus color (Cycowicz, Friedman, &
Snodgrass, 2001), amongst others.
Results from memory-disordered patients have also confirmed associations between successful episodic retrieval and late positive ERPs. Amnesic patients exhibited impaired conscious recognition as well as reduced or absent late positive amplitudes (e.g. Olichney et al.,
2006; e.g. Olichney et al., 2000). In addition, administration of benzodiazepine drugs to healthy subjects prior to encoding creates a temporary state of amnesia, and following this manipulation, both recollection and late parietal potentials are severely disrupted (e.g. Curran, DeBuse,
Woroch, & Hirshman, 2006). Taken together, evidence taken from a variety of experimental paradigms thus converges on the conclusion that recollection is a distinct expression of memory that reliably occurs with a particular ERP signature (Figure 4).
10. ERP Correlates of Post-Retrieval Processing
Another memory phenomenon may be indexed by positive potentials at prefrontal scalp locations beginning approximately 500 ms after stimulus onset and extending for up to several seconds. These potentials often display a right-sided distribution. They do not index retrieval 28 success, in that they tend to be similar for both successful and unsuccessful retrieval attempts.
Instead, these frontal potentials are thought to index effortful retrieval processing, manipulation
of working-memory contents, and/or post-decisional mnemonic processing such as further
evaluation.
The functional significance of these late frontal potentials was difficult to decipher, partly
due to the absence of a direct connection with retrieval success. Experimental manipulations of retrieval demands, however, were useful for clarifying the cognitive operations indexed by these potentials. In one study, images of common objects were encoded, and these objects were presented again at test in either the identical format or perceptually altered (Ranganath & Paller,
1999). Subjects performed one of two tests that differed in the demands placed on effortful
retrieval of perceptual detail. In the highly demanding test, subjects responded “old” only to
objects in the identical format whereas in the less-demanding test objects were to be endorsed as
old regardless of any format alterations. Late frontal potentials were larger in the highly
demanding test than in the less-demanding test for both categories of old as well as for new
stimuli (Figure 5). These potentials thus appeared to track retrieval effort as manipulated across
these two recognition tests. Subsequent investigations have confirmed this interpretation
(Leynes, 2002; Ranganath, Johnson, & D'Esposito, 2000; Ranganath & Paller, 2000), supporting
the view that late frontal potentials track strategic processing that accompanies retrieval, a
process likely mediated by prefrontal cortex. Hemispheric loci of late frontal potentials can also
vary as a function of retrieval demands, with a right-to-left shift accompanying an increase in the
complexity of the retrieved information (M. K. Johnson, Kounios, & Nolde, 1997; Nolde,
Johnson, & Raye, 1998).
29 11. Difficulties Identifying ERP Correlates of “Pure” Familiarity
Attempts to associate the memory experience of familiarity with specific ERPs have been
most controversial. Results from many studies have been taken to indicate that familiarity is
generically indexed by a negative potentials peaking approximately 400 ms post-stimulus, an
N400-like potential with reduced amplitudes (i.e., more positive ERPs) for old compared to new
items, especially at anterior locations. Beginning with the work of Düzel and colleagues (1997), many researchers have proposed that this frontal N400 old/new effect (FN400) indexes recognition with familiarity, in contradistinction to recollection (reviewed in Curran, Tepe, &
Piatt, 2006). Whereas late posterior potentials are greater in circumstances in which recollection is greater (for example, following semantically-deep vs. semantically-shallow encoding tasks), the frontal N400 old/new effect is generally found to be insensitive to such manipulations (e.g.
M. D. Rugg, Mark et al., 1998; M. D. Rugg, Walla et al., 1998). In addition, FN400 potentials have been associated with phenomenological familiarity as indexed by the remember/know paradigm (e.g. Duzel et al., 1997; Woodruff, Hayama, & Rugg, 2006) and have been found not to scale with the amount of recollection during tests of source memory (e.g. Wilding, 2000). In sum, FN400 potentials are sensitive to recognition success but not to manipulations that effect recollection, and thus are widely described in the extant ERP literature as a general correlate of familiarity.
On the other hand, there is reason to doubt this generalization about FN400 potentials, because the majority of findings used to argue in favor of this association are indirect (see below and Paller et al., in press, for a review). Specifically, the logic of interpretation has generally been that FN400 potentials reflect recognition memory and do not behave as a neural correlated of recollection, and so the inference made has been that they therefore index familiarity. A direct 30 challenge to this fragile interpretation arose with the identification of intact N400 repetition
effects in amnesic patients which encompassed frontal electrodes at which FN400 potentials are
commonly observed (Olichney et al., 2000). A reasonable generalization is that amnesia disrupts
familiarity, in the sense that a patient with severe amnesia does not behave as if people and
various objects they encounter feel familiar. This generalization also stands on sound empirical
footing (Knowlton & Squire, 1995; Yonelinas, Kroll, Dobbins, Lazzara, & Knight, 1998). Thus,
one might expect patients with amnesia to exhibit reduced FN400 old/new effects if these effects
indeed index familiarity. Olichney and colleagues (2000) thus suggested that frontal FN400
potentials might instead reflect the operation of conceptual implicit memory processes (described
in Section 15) that can be engaged incidentally during recognition testing. In general, special
steps are necessary to isolate the contribution of separate but potentially co-occurring memory
phenomena to neural correlates of recognition memory, as discussed in detail in the next section.
Several studies have examined the phenomenon of familiarity without recollection using
faces, as in the classic example described by Mandler (1980) as the butcher-on-the-bus
phenomenon. The butcher’s face may seem extremely familiar yet not be identified when seen in
an unusual context, such as on the bus, whereas in the butcher’s shop, familiarity is more likely to occur together with memory for additional episodic and semantic information that uniquely identifies the specific person. Yovel and Paller (2004) used a variation of the remember/know
paradigm to segregate trials for separate analyses of recollection and familiarity. Recognition
with familiarity, compared to correct rejections of new faces, co-occurred with late positive
ERPs that were maximal at midline parietal locations; recognition with recollection co-occurred
with late positive ERPs that were much larger in amplitude, spanned a longer time interval, and
showed a slightly more anterior topography. MacKenzie and Donaldson (in press) conducted a 31 similar study and found a statistically significant topographic differences; familiarity was
associated with late posterior ERPs and recollection with larger and more anterior ERPs.
However, a third study (Curran & Hancock, in press) used a more heterogeneous mixture of
faces (i.e., faces with different racial and ethnic features) and failed to replicate this pattern;
instead attributing an FN400 effect to familiarity for faces. Of course, it is plausible that
characteristics of the people shown could influence the extent to which repeated faces engage
conceptual priming, although this idea deserves further study. In sum, most studies of familiarity
experiences during face processing associated familiarity with late posterior ERPs, not with the
earlier frontal FN400 potentials described in prior studies that used words or nameable objects.
Further studies are need to determine whether this divergence can be explained by showing that
familiarity entails different neural events for verbal versus facial stimuli, that heterogeneity of
face stimuli plays a crucial role, or whether alternative interpretations of FN400 potentials are viable.
Results from a limited number of studies can be taken as tentative evidence that familiarity may be indexed by potentials occurring earlier than FN400s (Curran & Cleary, 2003;
Diana, Vilberg, & Reder, 2005; Duarte, Ranganath, Winward, Hayward, & Knight, 2004;
Friedman, 2004; Tsivilis, Otten, & Rugg, 2001). These potentials occur between 100 and 300 ms
but otherwise closely resemble FN400 potentials – frontal ERPs to old items are more positive
than to new items. Like FN400 potentials, these earlier frontal potentials have been associated
with familiarity based on indirect evidence: the effect is present in association with
phenomenological reports of familiarity and does not scale with recollection. More evidence is
needed to determine if these potentials indeed index familiarity as opposed to other potentially 32 co-occurring memory phenomena, such as various forms of priming and the initiation of memory
search (e.g., Diana et al., 2005).
12. Using ERPs to Contrast Memory Subtypes
Memory performance undoubtedly reflects the operation of a variety of neural systems
(White, in press). Multiple memory systems or processes make variable contributions to
performance on different mnemonic tasks. ERP investigations are especially well-suited for
identifying the occurrence of these variable contributions and thereby disentangling the operation of distinct memory components. Indeed, we must first come to understand these separate components before we can work out how their interactions ultimately produce memory abilities.
One distinction that has been particularly amenable to investigation with ERPs is that between explicit memory and forms of implicit memory known collectively as priming. In an explicit memory test, specific reference is made to remembering information learned earlier. In an implicit test of memory, in contrast, no reference is made to learning episodes, but rather, memory is demonstrated via a change in performance in a certain task due to a prior event that may or may not be consciously remembered. Contrasts between these two broad categories of memory phenomena have been very prominent in memory research over the past two decades
(Grill-Spector, in press; Schacter, in press). Performance on explicit memory tests is typically disrupted in cases of amnesia, as described above. On the other hand, many types of implicit memory have been shown to be preserved in amnesia. Priming is a form of implicit memory that is indexed behaviorally as faster or more accurate responses on specialized priming tests, independent of conscious memory for study episodes. The most common types of priming tests are used to measure perceptual priming (also called item-specific implicit memory). These 33 behavioral effects are thought to reflect facilitated or more fluent perceptual processing of the physical features of repeated items, distinct from accessing a memory for the full episode in which the item occurred. A different set of mechanisms may be responsible for some types of priming (i.e., conceptual priming, novel-information priming, new-association priming, association-specific priming, or cross-domain priming), and in some of these cases priming may not be preserved in amnesia, although this is a topic currently under active investigation.
When memory tests are given to healthy individuals, performance may be guided by
explicit memory, implicit memory, or by some combination. In this sense, memory tests may not be “process-pure.” In addition to acknowledging that behavioral measures in memory tests can
reflect multiple memory processes, it is important to note that neural measures such as ERPs are
liable to be influenced by multiple memory processes as well. Moreover, neural measures can
reflect memory processes whether or not those processes influence behavioral performance. In
either implicit or explicit memory tests, ERPs can reflect neurocognitive processes responsible for both types of memory. Experimental manipulations that selectively influence the operation of
distinct components of memory are thus essential. Otherwise, ERP or other neuroimaging results
cannot be unequivocally associated with one type of memory versus the other.
13. Perceptual Priming and Recognition Memory
Isolating neural correlates of perceptual priming uncontaminated by those of conscious
remembering is problematic because of the difficulty of preventing subjects from recalling prior
episodes during priming tests. Similarly, the automatic processing that supports perceptual
priming may occur during recognition tests, even if behavioral measures of priming are not 34 obtained, and this processing can potentially be reflected in neural measures accompanying recognition.
In order to isolate ERP correlates of perceptual priming, Paller and colleagues (2003) used a condition in which faces were encoded only to a minimal extent such that priming occurred in the absence of recognition. Subjects viewed each face for 100 ms at a central location while simultaneously a yellow cross was shown unpredictably in one of the four quadrants 1.8˚ from fixation. While maintaining central fixation, subjects attempted to discriminate between two subtly different types of yellow crosses, and further stimulus processing was disrupted via backward masking. On a subsequent test, recognition of these minimally processed faces was not significantly better than chance. Perceptual priming for these faces, however, was observed behaviorally on two implicit memory tests. The logic of this design was thus that ERPs elicited by these faces could conceivably reflect neural events responsible for perceptual priming, whereas contributions from recognition processes would be negligible. Faces in another condition were presented for a longer duration, without disruptive perifoveal visual discriminations or backward masking, and were recognized at above-chance levels. These two conditions thus provided a direct comparison between ERPs associated with conscious memory for faces and ERPs associated with perceptual priming. Recognition-related neural correlates included late positive potentials (Figure 6a), closely resembling responses previously associated with face-cued recollection (Paller, 2000; Paller et al., 1999), whereas perceptual priming was associated with a relative ERP negativity over anterior recording electrodes from approximately 200-400 ms after face onset (Figure 6b). Spatiotemporally distinct
ERPs of opposite polarities were thus associated with conscious remembering versus perceptual priming. This pattern of neuroimaging findings complements neuroanatomical dissociations 35 identified in amnesic patients; the results imply that implicit access to memory is supported by
processing within a network of brain regions that is qualitatively distinct from that supporting
conscious access to memory.
Evidence for the independence of implicit and explicit memory can also be derived from
contrasts between neural correlates of encoding responsible for later perceptual priming versus
recollection. Schott and colleagues (B. Schott, Richardson-Klavehn, Heinze, & Duzel, 2002)
used deep/semantic vs. shallow/nonsemantic encoding conditions, followed by an ingenious,
two-stage procedure to assess memory. Three-letter word stems were presented in an explicit
memory test (i.e., cued recall), but subjects were encouraged to guess if they could not remember a studied word so that priming might also occur. After each stem was completed, subjects indicated using strict criteria whether they recognized the word from the encoding phase. Trials were categorized as showing priming if the subject produced the word at the completion stage but failed to endorse it as an old word (i.e., priming-without-recognition). Trials were categorized as remembered when the correct response was made at both stages, and as forgotten if not produced at the completion stage. Subsequent-memory analyses thus revealed a Dm for priming that took the form of a relative ERP negativity over central and fronto-central locations approximately 200-400 ms after word onset (resembling ERP correlates of perceptual priming identified during memory testing, e.g. Paller et al., 2003). Furthermore, Dm for priming was distinct from ERP differences between deep versus shallow encoding as well as from Dm for recognition, which both included relatively positive potentials at later intervals with different topographies. Collectively, these results (along with those from a follow-up study using fMRI, B.
H. Schott et al., 2006) constitute critical first steps in characterizing the neurocognitive 36 relationship between expressions of explicit memory and expressions of perceptual implicit memory.
14. Conceptual Priming
Another form of priming known as conceptual priming (Henson, 2003; Schacter &
Buckner, 1998) can occur whenever meaningful stimuli are repeated. Behavioral measures of conceptual priming are similar to those of perceptual priming in that they can occur in the absence of awareness of remembering and typically take the form of faster or more accurate responses to a specific stimulus. These alterations of behavioral responses are thought to reflect facilitated processing of stimulus meaning, and potentially support some of the short-term mnemonic operations that are preserved in amnesia, such as language comprehension.
Because the neural processing that supports conceptual priming can occur whenever meaningful stimuli are repeated, regardless of whether a behavioral test of conceptual priming is provided, it is possible that neural activity associated with conceptual priming occurs incidentally during tests of recognition memory for meaningful stimuli. As reviewed above,
FN400 potentials at retrieval have been postulated to index the explicit memory capability termed familiarity. The finding that similar potentials are intact in amnesic patients (Olichney et al., 2000), however, raised the possibility that FN400s instead reflect a form of memory that is not disrupted in amnesia. Olichney and colleagues (2000) proposed that residual conceptual priming in amnesic patients could be reflected by FN400 potentials. It is possible that FN400 potentials do not index familiarity but instead reflect conceptual priming that occurs concurrently with explicit memory (for a review, see below and Paller et al., in press). Indeed, conceptual 37 priming and familiarity are similar in many ways, and further work is needed to disentangle these two memory functions, and it is this distinction that is the centerpiece of the rest of this thesis.
38 Part Three: Existing Evidence against a Unique ERP Signature of Familiarity
15. Familiarity and Conceptual Priming
The experience of recollection has been convincingly associated with late-onset parietal
ERPs (reviewed in Section 9). Familiarity, on the other hand, has been only indirectly linked to
FN400 potentials (reviewed in Section 11). This evidence is evaluated in detail below, but can be summarized as follows: FN400 (i) is a neural correlate of recognition memory that is often found to be insensitive to experimental manipulations that are known to enhance recollection (e.g. encoding conditions that involve deep semantic processing), and (ii) is often associated with phenomenological descriptions of familiarity during recognition tests (i.e., when a subject recognizes that a stimulus is old but does not express memory for contextual detail of the initial
encounter). Thus, FN400 potentials have been attributed to familiarity due primarily to a process
of elimination – they are neural markers of recognition memory that do not index recollection,
and so seemingly must index familiarity.
This interpretation is extremely problematic because it is commonly accepted that multiple
types of memory can co-occur, thus making it difficult to disentangle the neural events
associated with each. In this case, neuroimaging measures collected during a recognition memory test can index explicit expressions of memory that support the ability to identify repeated stimuli, such as recollection and familiarity, but can also index other forms of memory,
such as conceptual priming.
Familiarity and conceptual priming overlap considerably in their sensitivity to different
experimental manipulations [e.g., disrupted to a similar extent following frontal lobe lesions 39 (Gershberg, 1997), administration of benzodiazepines (Bishop & Curran, 1998), and encoding conditions such as depth of processing (Hamann, 1990; Srinivas & Roediger, 1990), study duration (Challis & Sidhu, 1993), and divided attention (Mulligan, 1998; Mulligan & Stone,
1999); reviewed in (Yonelinas, 2002)]. Indeed, many researchers posit that conceptual priming directly supports the experience of familiarity, whereby fluent conceptual processing gives rise to the phenomenological experience of familiarity (Rajaram & Geraci, 2000; Verfaellie &
Cermak, 1999; Whittlesea & Williams, 1998; Wolk et al., 2005). Thus, it is extremely difficult to identify experimental manipulations suitable for disentangling familiarity from conceptual priming, as reviewed below in Part 3.
It is possible that FN400 potentials reflect familiarity, conceptual priming, or some unknown combination of these or other memory processes. Determining the functional significance of FN400 potentials will provide evidence critical to two current theoretical controversies.
Controversy 1: The functional relationship between familiarity and conceptual priming is ambiguous. Whereas several findings indicate that perceptual priming does not promote familiarity (Hamann & Squire, 1997; Stark & Squire, 2000; Wagner, Gabrieli, & Verfaellie,
1997), many researchers propose that conceptual priming supports familiarity (Jacoby & Dallas,
1981; Johnston, Dark, & Jacoby, 1985; Rajaram & Geraci, 2000; Verfaellie & Cermak, 1999;
Wagner et al., 1997; Wolk et al., 2005). Behavioral evidence in healthy individuals suggests that conceptual priming leads to a bias to indicate that an item has been seen before independent of if it has actually been seen before, thus indicating that veridical familiarity must be supported by processes other than conceptual priming (Rajaram & Geraci, 2000). 40 In amnesia, behavioral results often indicate that conceptual priming and explicit memory operate independently, in that patients exhibit preserved conceptual priming despite impaired recall of learning episodes (Graf, Squire, & Mandler, 1984; Keane et al., 1997; Levy, Stark, &
Squire, 2004; Vaidya, Gabrieli, Keane, & Monti, 1995). However, it is entirely possible that conceptual priming and explicit memory operate differently in intact brains, in which conceptual priming could influence intact explicit memory.
If conceptual priming and familiarity rely on common mechanisms, one might expect neural correlates of the two memory phenomena to coincide and overlap to a large extent. It is therefore critical to validate neural correlates of conceptual priming and familiarity in order to specify their functional relationship. Furthermore, it is reasonable to assume that this relationship can be best studies in intact brains, and thus neuroimaging could provide key insights beyond those available using solely behavioral measures.
Controversy 2: A dominant perspective in memory research is that recognition in both humans and other animals is supported by two memory processes termed recollection and familiarity (Aggleton & Brown, 2006; Yonelinas, 2002). These terms are deceptive in that they overlap considerably with the phenomenological experiences of recollection and familiarity that
I have already described. However, in the context of dual-process theory, recollection and familiarity are two distinct retrieval processes that produce the phenomenological experiences of recollection and familiarity, respectively. Recognition of a stimulus or event is thought to be supported by recollection when contextual details regarding the initial encounter are also retrieved. In contrast, familiarity-based recognition remains unsubstantiated by the retrieval of any pertinent detail, such as when a person’s name cannot be recalled yet his face nonetheless 41 seems familiar. This dual-process account of recognition memory has become a dominant
theoretical framework guiding research into the neural substrates of explicit memory.
Nonetheless, there is much debate over whether evidence in humans favors dual-process over single-process models, in which recognition is supported by a unidimensional memory strength variable that can encompass the phenomenological experiences of both recollection and familiarity (e.g. Wais, Wixted, Hopkins, & Squire, 2006; Wixted, in press). By this account, a single dimension of memory strength can produce the phenomenological experience of recollection when strong and the experience of familiarity when weak. Results from many neuroimaging experiments have supported the dual-process account by identifying distinct neural correlates of recollection and familiarity (reviewed in M.D. Rugg & Curran, in press; M. D.
Rugg & Yonelinas, 2003). In ERP studies, these include late posterior potentials for recollection and FN400 potentials for familiarity. Indeed, evidence in favor of dual-process models would be severely weakened if recollection and familiarity were found not to produce unique neural signatures.
It is thus critical that the functional significance of FN400 potentials be determined. If
FN400 potentials index familiarity and late-onset posterior potentials index recollection, as is commonly accepted, then there truly is a double dissociation between electrophysiological correlates of familiarity and recollection. If, however, FN400 potentials instead index conceptual priming, then the evidence does not support a dissociation and results from electrophysiological studies would thus support single-process models of explicit recognition.
The resolution of these controversies has specific ramifications for our understanding of the neurobiological substrates of recollection, familiarity, and conceptual priming. Computations performed by the hippocampus are widely held to support recollection, whereas the adjacent 42 neocortex of the parahippocampal gyrus is thought to be critical for familiarity (Aggleton and
Brown, 2006). These structures constitute the apex of sensory processing streams, and are thus positioned to integrate information into coherent memory representations. Although the exact operations performed are still poorly understood, some evidence indicates that the hippocampus and surrounding cortex perform unique computations that would be critical for long-term memory formation (Leutgeb, Leutgeb, Moser, & Moser, 2007). Nonetheless, some evidence indicates that the hippocampus supports both recollection and familiarity (e.g. Wixted and
Squire, 2004), and these two hypothesized processes cannot be readily dissociated based solely on evidence from the operations of medial temporal lobe structures.
Neural processing external to the medial temporal lobe could also differentially support recollection and familiarity. For instance, some evidence has indicated that two distinct retrieval processes are implemented in parietal cortex (reviewed in Wagner, Shannon, Kahn, & Buckner,
2005). It has yet to be determined if these processes map onto recollection and familiarity, or whether they do so via interactions with distinct medial temporal lobe structures. Determining if the recollection and familiarity have dissociable neural substrates is thus critical to assessing whether they represent fundamentally distinct expressions of memory. It is possible that the same neural computations performed in the same structures are critical for both memory processes, and that familiarity and recollection differ only in the extent to which these computations are engaged during a retrieval event.
Similar reasoning applies to determining the border between familiarity and conceptual priming. The neural mechanisms that support conceptual priming are poorly understood. Some evidence implicates processing within inferior frontal and anterior temporal cortex (e.g.
Donaldson, Peterson, & Buckner, 2001), although it is possible that additional structures are 43 involved. Priming is generally associated with reduced activity within these regions, which could represent repetition-induced facilitations in the efficacy of retrieval processes. Although
conceptual priming does not likely depend on processing within medial temporal lobe structures
(e.g. Levy, Stark, and Squire, 2004), familiarity and conceptual priming could nonetheless
represent different behavioral manifestations of the same neural computations. Because
behavioral measures of conceptual priming and familiarity are often highly correlated (see
above), this proposal can only be assessed by examining their neural relationship.
16. Variables That Cannot Differentiate Familiarity from Conceptual Priming
As noted above, many experimental variables influence familiarity and conceptual
priming in a parallel fashion. Accordingly, much of the existing evidence does not favor the
hypothesis that FN400 potentials index familiarity over the hypothesis that FN400 potentials
index conceptual priming. Even though valid behavioral dissociations between familiarity and
conceptual priming are generally lacking in the ERP literature, experimental results have
nonetheless erroneously been taken as evidence that FN400 reflects familiarity. The review
provided here of studies that do not effectively dissociate conceptual priming and familiarity
calls attention to the serious limitations faced when attempting to infer the functional significance of FN400 potentials.
Words with a different plurality at initial encoding and memory testing, relative to words repeated with a consistent plurality, are recognized with greatly decreased behavioral estimates of recollection, whereas behavioral estimates of familiarity are minimally effected (Hintzman &
Curran, 1994). Curran (2000) employed this manipulation and found that recollection-related
ERPs showed the expected behavioral outcome for recollection (original-plurality amplitude 44 greater than plurality-reversed greater than novel word amplitude for late-onset posterior-
centered potentials), whereas FN400 did not differ between original-plurality and plurality-
reversed words. FN400 was thus taken as a neural marker of familiarity. Although conceptual
priming was not measured in this study, plurality reversal should theoretically exert minimal
influence on the magnitude of conceptual priming. Reading a word in either plurality activates roughly equivalent semantic representations, and thus repetition would enhance implicit access to these representations regardless of plurality reversal. Indeed, altering surface characteristics of words, a manipulation that is analogous to reversing plurality, impacts recognition memory but not conceptual priming (Luce & Lyons, 1998). Thus, this manipulation does not differentiate between the two accounts because the behavioral outcome of familiarity and conceptual priming following plurality reversal is identical.
Resting on similar logic to that underlying the use of plurality manipulations, several experiments using perceptual transitions have attributed FN400 potentials to familiarity.
Perceptual transitions, such as mirror-reversal of an image between encoding and recognition testing, are thought to impact recollection to a greater extent than familiarity. Accordingly, some
studies indicate that such manipulations do not significantly impact FN400, whereas they do
impact recollection-related late-onset parietal potentials [mirror-reversal, (Curran & Cleary,
2003); auditory-visual modality shift, (Curran & Dien, 2003), but see (Curran, Schacter,
Johnson, & Spinks, 2001)]. However, perceptual transitions similarly should not influence conceptual priming: viewing a nameable object in any orientation, for instance, results in access
to similar semantic representations that can be facilitated with repetition. These patterns of
results thus do not differentiate between the two accounts. 45 It is important to note that some experiments employing perceptual transitions have
uncovered patterns of FN400 effects that cannot be readily accommodated by either the
conceptual priming or familiarity accounts (Curran et al., 2001; Groh-Bordin, Zimmer, &
Mecklinger, 2005). For instance, Groh-Bordin and colleagues (2005) recorded ERPs during a
recognition test for identical and mirror-reversed objects, and identified reliable FN400
potentials for the identical-format condition but not the mirror-reversed condition. These results raise the possibility that other factors, such as perceptual similarity, can contribute to FN400
effects either directly or via an indirect influence on familiarity, conceptual priming, or both.
Further interpretation of these results necessitates progress in elaborating cognitive processes
that are influenced by mirror reversal, and do not presently provide convincing evidence for or
against the familiarity or conceptual priming accounts of FN400.
Relative to semantically-shallow encoding tasks (e.g. judging if the first and last letters in
a word are in alphabetical order), semantically-deep encoding tasks (e.g. incorporating a to-be-
remembered word into a sentence) are thought to engender stronger recollection and familiarity
during subsequent recognition tests (Toth, 1996). Manipulations of depth of encoding, however,
are unsatisfactory for mediating between the familiarity and conceptual priming accounts of
FN400, because conceptual priming has also been found to increase with encoding depth
(Hamann, 1990; Srinivas & Roediger, 1990). Furthermore, there is much ambiguity concerning the effect of depth of encoding on FN400. Ullsperger and colleagues (2000) identified FN400 effects for both deeply and shallowly encoded words, but direct comparisons of the effect’s magnitude were not made between these conditions. FN400 potentials identified in one study
(M. D. Rugg, Allan, & Birch, 2000) were sensitive to depth of encoding processing (present for deeply-encoded words, absent for shallowly-encoded words), whereas this difference was not 46 identified in another study [present for both deeply and shallowly-encoded words (M. D. Rugg,
Mark et al., 1998)].
One possible explanation for these anomalous results is that the structure of encoding and test sessions varied across experiments. Experiments that have identified FN400 effects for shallowly-encoded words (M. D. Rugg, Walla et al., 1998; Ullsperger et al., 2000) have employed a randomized study/test structure, whereby words studied under both conditions were randomly intermixed at both study and test, with appropriate cues preceding each study-phase word. In contrast, no FN400 effects for shallowly-encoded words were identified when depth conditions were blocked [i.e., one set of words studied deeply and then tested, followed by another set of words studied shallowly and then tested (M. D. Rugg et al., 2000)]. The use of blocked vs. randomized encoding conditions does not have clear relationship to variations in either familiarity or conceptual priming and, thus, this pattern of results can not add weight to either account of FN400. Likely, the failure to identify FN400 effects for shallowly-encoded words (M. D. Rugg et al., 2000) arose artificially as a result of the blocked design. ERPs correlates of recognition for shallowly-encoded words were measured relative to a baseline condition (correctly-rejected new words) that clearly differed between the tests following either deep or shallow encoding blocks. Differences in this baseline condition might have led to the failure to find an effect in the shallow condition. Taken together, this pattern of results thus suggests that FN400 amplitude does increase with encoding depth, but this manipulation does not distinguish between the two accounts.
In experiments that examine memory for source, items are associated with a particular experimental context when encoded (for example, words are spoken by either a male or female voice), and subsequent memory tests differentiate between correctly recognized old items for 47 which this source is either also correctly remembered (source-correct) or forgotten (source-
incorrect). Recognition of source-correct items requires a contribution from recollection whereas
recognition of source-incorrect items can be supported solely by familiarity. FN400 effects have
been identified that were identical for source-correct and source-incorrect items (Curran, 2000;
Trott et al., 1999) and that did not increase with increasing amounts of recollected source information (Wilding, 2000). In contrast, the amplitude of recollection-related late-onset parietal
potentials did increase with the quantity of recollected source information. These results have
thus been taken to indicate that FN400 indexes familiarity. However, it is reasonable to assume
that conceptual priming would also be unaffected by the amount of recollected source-specifying information. If behavioral indices of familiarity and conceptual priming are both unaffected by the quantity of recollected information, these patterns of results do not favor one hypothesis over the other.
A similar line of reasoning extends to experiments that examine memory when the encoding context is reinstated during the test phase vs. when the context is altered. For instance, words can be paired at encoding and then presented again either in the same word pair or paired
with different words (Trott et al., 1999). Contextual reinstatement facilitates remembering,
although it is unclear whether this advantage is due to an increase in recollection, familiarity, or
both. It is also conceivable that contextual reinstatement enhances conceptual priming, given that
congruent context provides cues to additional conceptual information that can be facilitated with
repetition. Contextual reinstatement increases the amplitude of FN400 potentials compared to
repetition with different context (Olichney et al., 2000; Trott et al., 1999), and these results
cannot differentiate between the two accounts given the uncertainty of whether familiarity,
conceptual priming, or both are enhanced by reinstatement. 48 Subsequent recognition memory can be reliably weakened by dividing attention during
encoding. For example, while subjects attempt to encode visual materials, a spoken digit can be
presented during every trial and subjects can be required to indicate whether the digit presented
several trials past was odd or even. Divided attention leads to a substantial decrease in
subsequent recollection, and impacts subsequent familiarity to a lesser degree (Yonelinas, 2001).
Behavioral evidence has indicated that conceptual priming, like familiarity, is weakened to a
minimal degree by divided attention at encoding (Mulligan, 1998; Mulligan & Stone, 1999).
However, it is also possible that divided attention at encoding does not impact conceptual
priming based on the following scenario. With full attention at encoding, conceptual priming
performance may be contaminated by explicit memory, leading to an artificial boost in
conceptual priming for the full-attention control conditions and an apparent decrement in
conceptual priming for divided-attention conditions. Taken together, dividing attention during encoding exerts either a weak influence on both familiarity and conceptual priming, or a weak influence on familiarity and no influence on conceptual priming.
Curran (2004) divided attention at encoding and found a substantial decrement (relative to full attention) in subsequent behavioral measures of recollection and a weaker, yet reliable, decrement in subsequent behavioral measures of familiarity. Late-onset parietal potential amplitudes were also reduced, whereas FN400 was either equivalent following full and divided attention (Experiment 1), or uniquely present following full attention (Experiment 2). Although these contradictory findings were taken to support the familiarity account, it is actually impossible to ascertain the relevance to either of the two accounts because the absence of behavioral measures of conceptual priming in this study weakens any interpretations. If conceptual priming was not impacted by dividing attention, then the findings of Experiment 1 49 would support the conceptual priming account because FN400 was also not impacted whereas
behavioral estimates of familiarity were impacted. However, if this were the case, the findings of
Experiment 2 would then, contradictorily, support the familiarity hypothesis. If instead
conceptual priming was impacted by dividing attention, just as were behavioral estimates of
familiarity, then the results of Experiment 2 could not distinguish between the two accounts and,
furthermore, the findings of Experiment 1 would speak against both accounts. Clearly, behavioral measures of both familiarity and conceptual priming are required in order to
disambiguate the relevance of divided attention to the two hypotheses.
Administering benzodiazepines, such as midazolam, during encoding severely impacts
performance on recognition memory tests given after the drug has cleared the system. This
temporary amnesia greatly reduces recollection and also consistently reduces familiarity to a
lesser extent (Hirshman, Fisher, Henthorn, Arndt, & Passannante, 2002; Mintzer & Griffiths,
2000). Curran et al. (2006) administered midazolam during encoding and found that subsequent
recognition was impaired relative to a saline-control. FN400 was equivalent for midazolam and
saline-control encoding conditions. In contrast, late-onset parietal ERPs were significantly
reduced following administration of midazolam relative to saline. Based on this finding, FN400
was attributed to familiarity. This interpretation, however, was unwarranted given that
midazolam consistently impacts familiarity, albeit to a minimal degree, whereas FN400 amplitude was not impacted. The status of conceptual priming following midazolam treatment has not been determined, although an experiment using a similar benzodiazepine (lorazepam) did find a small reduction relative to saline (Bishop & Curran, 1998). However, as is the case for dividing attention, this effect could be artificial in that conceptual priming could have been
relatively higher following saline encoding due to the use of explicit strategies that were not 50 available following benzodiazepine treatment. The relevance of Curran and colleagues’ (2006)
findings to the familiarity and conceptual priming accounts is thus ambiguous. If conceptual
priming was intact following midzolam encoding, then results would support the conceptual
priming account. If, however, both familiarity and conceptual priming were reduced by the drug,
then the use of midazolam and these ERP results cannot distinguish between the familiarity and
conceptual priming accounts.
17. Variables That Can Potentially Differentiate Familiarity from Conceptual Priming
As noted above, some experimental variables exert relatively differential effects on familiarity and conceptual priming. Therefore, these variables provide promise to adjudicate between the two accounts. Again, it is important to note that proposed differences in the effects of these variables on familiarity and conceptual priming stem primarily from prior behavioral studies and entirely theoretical hypotheses. Far stronger evidence in favor of either account would require that behavioral estimates of both memory processes be obtained under similar experimental conditions. Nonetheless, I will now discuss these variables in turn.
Explicit tests of memory cue subjects to intentionally retrieve information from episodic memory whereas implicit memory tests are designed to tap incidental expression of memory that can occur without retrieval intention. Because retrieval intention involves the active search of the contents of memory, it is more likely that information will be successfully retrieved during explicit compared to implicit memory tests. Therefore, familiarity should be stronger during explicit compared to implicit memory tests. Conceptual priming, on the other hand, does not require the intention to retrieve, but should occur more-or-less “automatically” as a result of prior conceptual processing (provided that sufficient attention is directed to stimuli such that 51 conceptual information may be extracted). Whereas the differential influence of retrieval
intention on familiarity and conceptual priming is theoretically tenable, it is important to note
that this assertion is difficult to test empirically, given that it is problematic to index either
priming during explicit (direct) tests or familiarity during implicit (indirect) tests.
The effect of retrieval intention FN400 is unclear. Results from some studies (Curran,
1999; Curran, Tanaka, & Weiskopf, 2002) included identical FN400 effects in intentional and
incidental retrieval conditions whereas other studies (Groh-Bordin et al., 2005; Guillem, Bicu, &
Debruille, 2001; Nessler, Mecklinger, & Penney, 2005) found that FN400 was reliable in an
intentional condition but not an incidental condition. One experimental difference that potentially accounts for this anomalous pattern of results is that studies reporting that intention had no effect
varied retrieval intention within individual subjects whereas one report (Groh-Bordin et al.,
2005) of an effect of retrieval intention resulted when intention was manipulated across-subjects.
It is possible that subjects in both intentional and incidental retrieval conditions inadvertently
used intentional retrieval despite the instructions, whereas this confound was not problematic
when intentional or incidental retrieval instructions were mutually exclusive. This explanation,
however, is not in accord with all of the reported results [i.e., Guillem et al. (2001) and Nessler et
al. (2005) both varied intention within-subjects, and the mid-frontal old/new effect was found to
vary with intention], and it is thus premature to conclude whether influences of retrieval
intention on FN400 argue in favor of one hypothesis or the other.
Some stimuli carry with them high levels of conceptual or symbolic meaning (e.g., pictures
of known individuals and words) whereas some do not (e.g. abstract line drawings). Conceptual
priming is strongly influenced by stimulus meaning, whereby conceptual priming can occur for
meaningful stimuli to a larger extent than for meaningless stimuli. On the other hand, stimuli can 52 be recognized with familiarity largely independent of the degree of inherent conceptual meaning.
Thus, the two hypotheses make different FN400 predictions for stimuli of high vs. low inherent
meaningfulness. The familiarity account asserts that FN400 should not vary with the degree of
inherent meaning, whereas the conceptual priming account asserts that only stimuli of high
inherent meaning should elicit a FN400.
Unfortunately, the empirical evidence is mixed. In some studies, FN400 potentials have
been found to be unchanged by the degree of inherent meaning [e.g., present for both words and
pseudowords (Curran, 1999)]. Moreover, stimuli with extremely low levels of inherent meaning,
such as blobs, geometric shapes, and unfamiliar faces, often elicit reliable FN400 potentials
(Curran et al., 2002; Groh-Bordin, Zimmer, & Ecker, 2006; Guillem et al., 2001; Nessler et al.,
2005; Penney, Mecklinger, & Nessler, 2001). In contrast, FN400 was not present for unfamiliar faces that were remembered solely on the basis of familiarity (MacKenzie & Donaldson, in press; Yovel & Paller, 2004). Results are thus not unequivocally in favor of either account. It is
important to note, however, that studies that employ stimuli with extremely low levels of
inherent meaning are subject to the potential confound that some meaningless stimuli are indeed meaningful to some subjects, and could therefore support conceptual priming. For instance,
Curran and colleagues (2002) reported that some subjects found that a purportedly meaningless blob “looked like Texas” on a map. It is thus possible that the majority of evidence is actually in favor of the conceptual priming account, and this possibility is explored empirically below in
Part Six and Part Seven.
In directed forgetting paradigms, encoded items are followed by either a cue to remember or a cue to forget, and subsequent memory performance is hindered for actively-suppressed compared to actively-encoded material. This detrimental effect does not result merely from a 53 lack of intentional encoding, but is rather an effortful process that is thought to entail the active
suppression of encoded materials or disruption of working memory maintenance processing, and can be neurally dissociated from other encoding manipulations, such as depth of semantic processing, that effect memory strength (Golding & MacLeod, 1998; Reber et al., 2002;
Ullsperger et al., 2000). Given that both to-be-remembered and to-be-forgotten items are encoded to an identical extent before delivery of the cue, it is theoretically plausible that directed forgetting disrupts explicit memory processes that depend on active maintenance, including both recollection and familiarity, whereas conceptual priming would be influenced to a minimal degree given that semantic information necessary to support subsequent priming would be accessed prior to the cue. Ullsperger and colleagues (2000) found that FN400 was reduced for to- be-remembered and to-be-forgotten words. Thus, FN400 behaved as an index of familiarity.
However, a stronger interpretation that differentially favors either the familiarity or conceptual priming hypothesis will require directly assessing the influence of directed forgetting on behavioral indices of both familiarity and conceptual priming.
18. Direct Tests that have Failed to Validate FN400 Correlates of Familiarity
Two recent experiments have attempted to directly assess the familiarity hypothesis.
Despite the theoretical weaknesses described below, both have been taken as direct evidence in support of the hypothesis that FN400 indexes familiarity.
Rugg and colleagues (Woodruff et al., 2006) utilized phenomenological estimates of recollection and familiarity in concert with confidence decisions in order to directly assess the relationship between familiarity strength and FN400. In this experiment, subjects discriminated novel from repeated common words and introspectively classified their recognition experience as 54 ‘remember’ if recognition was substantiated by retrieval of specific encoding detail and ‘know’ if no details were simultaneously retrieved. Items that garnered ‘know’ responses were simultaneously given confidence ratings on a four point scale, ranging from ‘confident old’ to
‘confident new’, such that the familiarity strength of each item was tracked. FN400 amplitude varied directly with familiarity confidence, with the largest ERP amplitude difference for the
‘confident old’ category, and the smallest difference for the ‘confident new’ category. The authors thus concluded that FN400 is a direct index of familiarity.
A fatal methodological flaw, however, prevents these results from supporting the familiarity account. Subjects were quite accurate at discriminating old from new items using the familiarity confidence scale and, accordingly, old items predominated in the high confidence response category and diminished as confidence decreased. Accordingly, new items were scarce in the high confidence response category and increased as confidence decreased. Due to low stimulus counts in some of the response categories, both old items and new items were collapsed together to form ERP averages for each level of familiarity confidence. Because FN400 is greater for old compared to new items, the graded amplitude differences observed across confidence levels could have arisen artificially due to the graded number of old items across confidence levels. [an identical study that employed fMRI (Yonelinas, Otten, Shaw, & Rugg,
2005) is subject to the same limitation].
In an attempt to circumvent this shortcoming, an auxiliary analysis of FN400 amplitude was conducted for old items and new items with purportedly equated levels of familiarity confidence. To do so, a subset of old and new trials from each confidence level were randomly selected such that, overall, the number of trials from each confidence level was equated. When familiarity was equated using this method, FN400 potentials were not identified. The authors 55 thus concluded that the FN400 effect that was graded with confidence did not merely reflect a
diluted old/new effect. The problem with this approach is three-fold. It is likely that familiarity
confidence is not equated using this approach, given that response error and guessing could have resulted in the very small number of low-confidence old items and high-confidence new items, and these trials would then not reflect the same level of familiarity confidence as trials for which confidence ratings veridically represented old/new status. Indeed, the old and new averages were biased to include a greater number of low-confidence trials, given the small number of high- confidence “new” responses. Furthermore, the overall level of familiarity confidence captured by the old and new averages varied highly across subjects, as indicated by the highly variable trial counts that were reported, and the failure to find an old/new difference could be due to low power in many subjects. Finally, this result was obtained by a single random selection of trials, and it is thus difficult to determine if the results obtained would persist if a different random set
of trials were selected.
Even if we assume that FN400 did indeed vary monotonically with familiarity
confidence, this evidence would still fall short of unequivocally supporting the familiarity
hypothesis. One cannot assume that conceptual priming magnitude was orthogonal to familiarity
confidence in that the same factors responsible for variations in familiarity confidence across
trials – though they were not measured – could have also led to parallel variations in conceptual
priming magnitude. Because the two memory phenomena respond similarly to various experimental variables, as outlined above, the evidence for attributing FN400 to familiarity was no stronger than the evidence for attributing FN400 potentials to conceptual priming. Thus, the
fundamental limitation of this study, as with previous studies, was the absence of measures of
conceptual priming. 56 Another study (Azimian-Faridani & Wilding, 2006) employed a method known as
“criterion shift” in an attempt to identify electrophysiological correlates of familiarity. Subjects
performed recognition memory tests for common words using either a conservative or liberal
criterion for making an “old” response. Recognition was thus hypothesized to be based on
stronger familiarity in the conservative condition than the liberal condition, whereas conceptual
priming was matched. FN400 potentials were more positive for old compared to new items and
adopting a conservative criterion led to more positive FN400 potentials for both old and new
items than did adopting a liberal criterion. Results were thus taken to indicate that FN400
potentials reflected familiarity. An alternative interpretation, however, is that processing that
strengthened explicit memory also enhanced conceptual priming. Again, it is insufficient to
assume that familiarity and conceptual priming are orthogonal without providing corresponding
behavioral evidence. Thus, results obtained using the criterion shift method do not unequivocally
support the familiarity hypothesis.
In conclusion, the extant literature cannot be used to infer the functional significance of
FN400 potentials. Furthermore, the most direct tests of the familiarity hypothesis did not provide
strong evidence that FN400 can be attributed to familiarity. A prime weakness in many studies has been the failure to account for conceptual priming in addition to familiarity by providing behavioral estimates of conceptual priming under circumstances similar to those used to obtain neural correlates of familiarity. By providing such measures, direct empirical evidence could be used to disambiguate the observed pattern of neuroimaging results. It is this tact that is taken in the following four experiments, the first of which uses behavioral estimates of conceptual priming and familiarity for famous faces to map corresponding electrophysiological signatures.
57 Part Four: ERP Correlates of Familiarity and Conceptual Priming for Famous Faces
19. Rationale
Previous experiments examining familiarity that were reviewed in Part 3 relied primarily on hypothesized relationships between experimental variables or manipulations and critical memory events. To provide much stronger relationships between neuroimaging measures of critical memory events, including familiarity and conceptual priming, it is necessary to collect relevant behavioral measures such that brain/behavior relationships can be substantiated. The current and forthcoming experiments take exactly this tact, and are thus poised to provide pertinent evidence regarding the neural foundations of familiarity and conceptual priming.
In this experiment, behavioral and neural correlates of familiarity and conceptual priming were sought. Because neural correlates of familiarity were found not to include FN400 potentials, but instead resemble smaller versions of recollection-related parietal-centered potentials (Yovel and Paller, 2004), it was of interest to obtain ERP correlates of conceptual priming for faces. Famous faces were selected because it was reasoned that associated biographical information could be conceptually primed, whereas it would be difficult to produce conceptual priming for nonfamous faces.
In an initial portion of the experiment, the potential for conceptual priming was manipulated selectively for a set of famous faces by presenting them along with relevant biographical information. Behavioral and neural correlates of conceptual priming were then collected during a subsequent portion of the experiment. Results from two control experiments were used to validate the conceptual nature of the behavioral priming effects. Following the test 58 of conceptual priming, ratings that captured variations in explicit memory were collected for
each famous face, such that neural correlates of explicit memory that occurred incidentally
during the priming test could also be measured. Thus, neural correlates of explicit memory
(which resembled those of familiarity, see below) and conceptual priming were obtained during the performance of a single behavioral task. This paradigm was advantageous in that it permitted a direct comparison of the neural signatures of familiarity and conceptual priming, without
introducing confounding influences of between-task comparisons.
20. Methods
For Experiment 1, behavioral and ERP data were collected from 10 right-handed native
English-speakers (6 female, ages 18-19 yrs). ERP data recorded from an additional three subjects
were excluded due to excessive eye-movement artifacts such that too few uncontaminated trials
were available. The pattern of behavioral results for the 10 subjects contributing ERP data was
identical to that for the entire group of 13 subjects. Behavioral data are thus reported for the
entire group.
Facial stimuli consisted of 180 photographs of celebrities (actors, politicians, musicians,
professional athletes, and TV personalities) and 180 photographs of nonfamous individuals. The
format of famous and nonfamous faces was similar and included only the head, in grayscale, on a
black background. Famous and nonfamous faces were divided into two sets of 90 (mean percent
female in each set = 50.3%).
Three biographical cues were compiled for each famous individual. The name was
always used as a cue, as well as two other short identifying pieces of information. Additional 59 cues most often included the title of a film or song, a television role, or a political office, and are provided in Appendix 1.
For each participant, one of the two sets of 90 famous faces was used for the condition designated the primed condition (see below). Participants were exposed to biographical cues for those 90 celebrities, and did not see the biographical cues for the other 90 celebrities. The set of celebrities assigned to the primed condition was counterbalanced across ERP subjects.
The experiment was comprised of three distinct phases. During the first phase, participants viewed faces belonging to three conditions: primed, unprimed, and nonfamous. The goal was for participants to think about person-specific information primarily for the primed celebrities. Biographical information was shown on the screen just prior to each face.
Participants indicated via button press whether each face matched the biographical cue that preceded it (Figure 7). Primed faces were 90 celebrity photographs preceded by a matching biographical cue. Unprimed faces were 90 celebrity photographs preceded by a biographical cue for a randomly selected primed face. Each of 90 nonfamous faces was also preceded by a biographical cue for a randomly selected primed face. In the unprimed and nonfamous conditions, mismatching name cues did not necessarily match on gender. The priming phase was divided into three segments such that each of the 270 faces was presented once per segment, each time with a different biographical cue. The specific information pertaining to each primed celebrity appeared 9 times (once per segment with the matching celebrity, once per segment with celebrity faces from the unprimed condition, and once per segment with nonfamous faces). Faces were shown in random order at a fast rate, as shown in Figure 7. This rapid presentation format, along with the task requirement to maintain information regarding a primed celebrity while evaluating each face and producing a response, functioned to limit the recall of information 60 related to unprimed celebrities. This procedure thus provoked subjects to bring to mind a greater amount of conceptual information regarding primed than unprimed individuals, allowing us to obtain measures of conceptual priming in the next phase of the experiment.
Approximately 5 min after phase 1, each of the 270 faces was shown again along with 90 nonfamous faces never seen before (to equate the number of famous and nonfamous faces) during phase 2. Subjects made a speeded go/no-go response by pressing a button as quickly as possible after each famous face (not pressing any button in response to nonfamous faces). Faces were presented in random order. Given that deciding whether a face is famous entails accessing pertinent conceptual information, we expected that responses to famous faces that were recently conceptually primed would be facilitated relative to those that were not.
The assessment of explicit memory was made during phase 3, and followed phase 1 by approximately 25 min. Stimuli and presentation parameters were identical to those in phase 2 except that stimuli were presented in a different random order. For each famous and nonfamous face, subjects made a rating using a 5-point scale where “1” corresponded to “very familiar” and
“5” corresponded to “not familiar at all.” This test can be said to measure lifetime familiarity, as assessed in a recent study of explicit memory (Diana et al., 2005), in that subjects were instructed to quickly make a gut-level memory assessment for which the source of the memory was irrelevant (i.e., it could include their experiences prior to the experiment or experiences during the experiment). This probe was designed to determine the extent to which each famous face was known to each participant while potentially remaining sensitive to the influence of previous phases of the experiment, thus yielding a behavioral index of the relative amount of explicit memory of any type occurring in response to each face. Given that ERP measures are potentials sensitive to any type of retrieval that takes place, this somewhat general measure of 61 explicit memory (including both episodic and semantic retrieval) is useful for also being sensitive to these multiple influences of memory on face processing.
Electroencephalographic recordings were made during phase 2 from 59 scalp sites using tin electrodes embedded in an elastic cap at locations designed to provide fairly even coverage across the scalp. Four channels were used for monitoring horizontal and vertical eye movements, and trials contaminated by electro-ocular artifacts were excluded from ERP analyses. Impedance was lowered to 5 kΩ or less. EEG signals were collected with a band pass of 0.05 to 200 Hz, sampled at a rate of 1000 Hz, and rereferenced offline to average mastoids. Each averaging epoch lasted 1200 ms, including 100 ms prior to stimulus onset. Baseline correction was performed by subtracting mean amplitude over the pre-stimulus interval from every post- stimulus timepoint.
ERPs elicited by famous faces during phase 2 were averaged in two different ways: (1)
ERPs were computed as a function of conceptual priming based on the presentation of biographical information during phase 1 (i.e., primed vs. unprimed conditions); (2) ERPs were computed as a function of subsequent ratings during phase 3, with faces endorsed with a high familiarity rating (“1” or “2” on the 5-point scale) operationally defined as high in explicit memory retrieval (HEM) and those endorsed with a low rating (“3”, “4”, or “5”) defined as low in explicit memory retrieval (LEM). Responses to famous faces that were not endorsed as famous in phase 2 (10% on average) were excluded from both analyses.
ERP waveforms included data from 10 subjects for the primed/unprimed contrast, and data from 8 subjects for the HEM/LEM contrast. Two subjects were excluded from the latter analysis because nearly all celebrities were endorsed with high ratings during phase 3, such that there were too few trials (<20) in the LEM condition. Data from a total of 8 subjects were 62 suitable for comparisons of primed and unprimed faces matched in explicit memory and of HEM and LEM faces matched in priming.
Formal statistical comparisons were performed on ERP waveforms derived from two different electro-ocular artifact identification procedures. The standard approach was to reject trials with artifacts in the -100 to 1100 ms range (23% of trials on average, SEM = 0.04%).
Formal statistical comparisons focused on the first 750 ms post-stimulus, whereas many of the artifacts occurred after this interval. Therefore, a supplementary analysis identified artifacts within the first 750 ms of each trial, such that only 9% of trials were rejected. This supplementary analysis yielded an identical pattern of results as the main analysis (i.e. the same null hypotheses were rejected but with different levels of significance), and so statistical results from the main analysis are emphasized, and artifact-free ERP waveforms for the entire epoch are presented.
Significant differences in ERP measurements were evaluated using repeated measures
ANOVA (α=0.05) with Geisser-Greenhouse corrections when necessary. Post-hoc pairwise comparisons were reported only if significant following Bonferroni correction. Timecourse analyses were conducted using consecutive 5-ms averaging windows with uncorrected pairwise comparisons between corresponding windows from each condition. For presentation purposes only, waveforms were smoothed with a 55-Hz lowpass zero-phase-shift Butterworth filter.
Behavioral data were collected from six subjects in Experiment 2 (three female, ages 18-
20) and six subjects in Experiment 3 (four female, ages 18-19). Except for the following modifications, Experiments 2 and 3 (behavioral control experiments) were conducted exactly as described above. In both experiments, EEG recordings were not made. Also, in both experiments there were changes for the unprimed condition in phase 1. In Experiment 2, strings of 63 meaningless characters (consonants and punctuation) preceded unprimed faces. Subjects responded to each face using one of three buttons to indicate if the preceding information matched, did not match, or was neutral (meaningless characters). “Neutral” responses were made by pressing the appropriate response button three times while counting backwards from three, such that recall of information related to unprimed celebrities was minimized by taxing working memory, as accomplished in Experiment 1 using biographical information. In phase 1 of
Experiment 3, an appropriate gender description (“male” or “female”) preceded every unprimed face. Half of the nonfamous faces were preceded by an inappropriate gender description instead of a celebrity fact. Subjects responded “match” (all primed and unprimed) or “nonmatch” (only nonfamous faces). All gender responses were made three times while counting backwards, such that working memory during unprimed faces was taxed to a similar extent in all 3 experiments.
21. Results
Conceptual priming was found both in speed and accuracy of button-press responses in phase 2. A high proportion of the celebrity faces were endorsed as famous during the priming test (90%, SE=1.85%). On average, reaction times on correct trials were 32 ms faster for primed faces than for unprimed faces [t(12)=3.08, p= 0.01, mean RT=620 and 652 ms, respectively].
Every subject exhibited reaction-time priming (range=4-71 ms). In addition, all subjects achieved higher accuracy for primed than unprimed faces [t(12)=4.96, p<0.001; mean hit rate
93% and 86%, respectively]. Subjects responded incorrectly to nonfamous faces very infrequently, and these false alarms occurred equally often for nonfamous faces presented for the first time during phase 2 (mean=5.7%) and nonfamous faces repeated from phase 1 64 (mean=5.7%). Behavioral responding for these two types of nonfamous faces also did not differ
in phase 3 [t(12)=1.41, p=0.18)] and so are considered together in all other analyses.
In related priming studies, Dobbins and colleagues (2004) showed that response learning
can contribute to priming effects. Here, a relative facilitation for responding “yes” in phase 2 for
primed compared to unprimed faces could theoretically have resulted due to the fact that primed
faces received “match” responses in phase 1 whereas unprimed faces received “nonmatch”
responses. Results from Experiments 2 and 3 ruled out this possibility, because the same
magnitude of priming was observed using different response requirements in these two
behavioral control experiments. In both designs, primed and unprimed faces received the same
response in phase 1, and reaction-time priming paralleling that in Experiment 1 was observed
[Experiment 2: mean priming=29 ms, t(5)=2.48, p=0.0557; Experiment 3: mean priming=27 ms,
t(5)=3.33, p=0.021].
Presentation of biographical information in phase 1 also influenced explicit memory
performance in phase 3; small differences in mean explicit memory ratings on the 5-point scale
were observed [2.09 vs. 2.37 for primed and unprimed faces, respectively, t(12)=5.63, p=0.001].
Nonfamous faces engaged very little explicit retrieval compared to celebrities [4.52 vs. 2.23 for
nonfamous and famous faces, respectively, t(12)=17.7, p<0.001]. The priming manipulation in
phase 1 can thus be said to have influenced both implicit and explicit memory; nevertheless, neural correlates of conceptual priming and explicit memory can be derived selectively by virtue
of analyses that take both types of behavioral memory measures into account.
An additional analysis established the feasibility of planned ERP comparisons. An ERP
contrast between priming and explicit memory would only be meaningful to the extent that
primed faces were not all rated as highly familiar and unprimed faces not all rated as less 65 familiar. Indeed, large numbers of both primed and unprimed faces were included in the HEM condition (an average of 56% primed faces and 44% unprimed faces) and likewise for the LEM condition (an average of 42% primed faces and 58% unprimed faces).
ERPs were found to be more positive for primed than unprimed famous faces from about
250-550 ms, and more positive for HEM than LEM faces from about 450-750 ms (Figure 8a).
ERPs were formally analyzed over two consecutive time intervals, 250-500 and 500-750 ms, and over three regions, defined by averaging waveforms from anterior, middle, and posterior scalp locations (Figure 8b). Differences in ERP amplitudes were analyzed using repeated-measures
ANOVA with factors: condition (primed/unprimed or HEM/LEM), region
(frontal/middle/posterior), and time interval (early 250-500 ms/late 500-750 ms). Comparing primed and unprimed faces yielded a significant main effect of condition [F(1,9)=10.04, p=0.01] and a three-way interaction [F(1.31,11.77)=12.12, p=0.003]. Comparing HEM and LEM faces also yielded a significant main effect of condition [F(1,7)=6.03, p=0.044] and a three-way interaction [F(1.09,7.63)=13.93, p=0.006]. Post-hoc pairwise comparisons between conditions, run separately for each region and interval and corrected for multiple comparisons, revealed significant differences between primed and unprimed famous faces early in the frontal region
[t(9)=3.81, p=0.004] and late in middle [t(9)=3.92, p=0.003] and posterior [t(9)=3.86, p=0.004] regions. In contrast, HEM/LEM differences were significant late in middle [t(7)=5.61, p<0.001] and posterior [t(7)=5.89, p<0.001] regions.
Figure 8b shows topographies of ERP differences between conditions averaged over the same two time intervals. Priming differences appeared in the early interval as a relative positivity maximal over frontal locations and in the late interval as a positivity over posterior locations.
Explicit memory differences appeared as a positivity localized to posterior locations and 66 maximal in the late interval. A direct test of the differential sensitivity of the early frontal
positivity to the two manipulations was conducted. Frontal amplitude differences over the 250-
500 ms interval were significantly greater for the priming contrast compared to the explicit
memory contrast [mean difference=1.26 µV vs. –0.05 µV, respectively, F(1,16)=4.64, p=0.04].
Given that exactly this kind of early frontal ERP difference has been hypothesized to
reflect conceptual priming (Yovel & Paller, 2004), we analyzed correlations between each subject’s reaction-time measure of conceptual priming and the amplitude of the early frontal positivity in the primed/unprimed contrast. ERPs were measured in each subject by selecting the electrode showing the greatest priming difference at 250-500 ms within the frontal region. An extremely strong correlation was found between this ERP measure and the magnitude of priming
[r2(8)=0.77, p<0.001]. In contrast, the maximum amplitude difference in the late interval
(measured at the location showing the largest difference at middle and posterior regions during
this interval) was not correlated with priming magnitude [r2(8)=0.03, p=0.64].
In a complementary analysis, we found that explicit memory was related to late posterior
ERP differences but not to the early frontal positivity. The mean difference in familiarity rating
was computed between primed and unprimed conditions. This behavioral measure of the
influence of the priming manipulation on explicit memory was marginally correlated with the
maximum primed/unprimed amplitude difference in the late interval at middle and posterior
regions [r2(8)=0.44, p=0.052], whereas it was not correlated with the early frontal positivity
[r2(8)=0.04, p=0.61].
ERP correlates of conceptual priming and explicit memory clearly differed in topography
(Figure 8b). This impression was substantiated by a significant condition-by-region interaction
[F(2,48)=5.14, p=0.001] in a comparison between ERPs averaged over each of the three regions 67 and subjected to amplitude normalization using the root-mean-square procedure (McCarthy &
Wood, 1985).
The timecourse of these effects was analyzed using consecutive 5-ms intervals for data from the frontal and posterior electrode locations in Figure 8a. Differences pertaining to priming were reliable at the frontal electrode primarily from 300-350 ms and again from 425-475 ms and at the posterior electrode from 425-475 ms. Differences pertaining to explicit memory were significant at the posterior electrode from approximately 425-710 ms. Therefore, the relatively large time intervals chosen for formal analyses effectively captured the between-condition differences.
Given that priming magnitude and familiarity ratings were not entirely independent, another analysis focused on differences due to either the priming manipulation or explicit memory with the other variable held constant. To this end, sets of primed and unprimed famous faces were identified that were matched in explicit memory (i.e., all highly familiar, rated with
“1” or “2” on the 5-point scale; 74% (SE=0.05%) of primed faces and 64% (SE=0.06%) of unprimed faces were given such ratings). Similarly, subsets of HEM and LEM famous faces were identified that were all unprimed. ERPs to primed and unprimed faces matched in explicit memory and to HEM and LEM faces matched in priming were thus computed (Figure 8c and d).
The difference between explicit-memory-matched primed and unprimed faces was a relative frontal positivity for primed from 300-550 ms. In contrast, the ERP difference between priming- matched HEM and LEM faces was a relative posterior positivity for HEM from 450-750 ms.
Pairwise comparisons indicated that the priming difference was significant early in the frontal region [t(7)=4.36, p=0.0033] whereas the explicit-memory difference was significant late in middle [t(7)=4.98, p=0.0016] and posterior [t(7)=4.31, p=0.0035] regions. 68 The procedures used here to elicit conceptual priming succeeded because subjects were
already knowledgeable about the biographical information presented. However, some of the
biographical cues were not person-specific, in that they could apply to several different
celebrities, and some were not known to some subjects. Nonetheless, the significant results found
using both behavioral and electrophysiological measures confirm that the procedure in phase 1
successfully prompted differential processing of conceptual information associated with primed
versus unprimed celebrities. Moreover, ERP priming effects could also be observed when short- lag priming was produced using conceptually related famous faces or names (Schweinberger,
Pickering, Burton, & Kaufmann, 2002), but priming at the delays used here entails different
memory processing than when the primed information is at the focus of attention when the target item appears.
Because ERPs associated with conceptual priming Experiment 1 were maximal over frontal electrodes, special consideration of possible electro-ocular artifacts is warranted. If this frontal ERP positivity was based on residual artifact, a relative negativity for the same contrast would be expected at electrodes positioned below each eye. Instead, ERPs at these electrodes were slightly more positive for primed compared to unprimed faces in the interval from 250-500 ms (0.41µV and 0.22µV at left and right electrodes, respectively). Thus, frontal ERP correlates of conceptual priming can be attributed to brain activity rather than electro-ocular artifact.
ERP correlates of explicit memory identified by the HEM/LEM contrast could index a combination of explicit memory processes, including recollection or familiarity for phase 1 episodes, retrieval of semantic information acquired prior to the experiment, and recollection or familiarity for relevant pre-experimental episodes. Accordingly, we cannot determine how much of the HEM/LEM contrast reflects pure familiarity. Neural responses to famous faces may 69 generally include retrieval beyond familiarity to the extent that people tend to recall biographical information when seeing the face. Indeed, when people view well-known celebrities, as in phase
2, recall of pre-experimental information may be virtually impossible to exclude. With nonfamous faces, however, pure familiarity experiences can be identified using variants of the
“remember”/”know” procedure (Gardiner & Java, 1991; Tulving, 1985), in which subjects introspect about their memory experiences to determine whether episodic information is recollected.
In a prior experiment (Yovel & Paller, 2004), we succeeded in characterizing neural correlates of pure familiarity with faces. Subjects first viewed novel faces presented with unique person-specific information (an occupation) in a study phase. Faces presented for recognition judgments in a test phase were endorsed with episodic recollection (remembering the face along with contextual or episodic information) or with episodic familiarity (endorsing the face as old but failing to remember the associated occupation and failing to recollect any specific episodic information). Figure 9 shows ERP waveforms and difference topographies associated with this experience of recognizing faces with pure familiarity, using the same format as in Figure 8 so as to allow a direct juxtaposition between ERP results of Yovel and Paller (2004) and ERP results from the present experiment. Importantly, ERP correlates of pure familiarity were nearly identical in timing and topography to ERP correlates of explicit memory, as observed in the
HEM/LEM contrast and, to a lesser extent, in the primed/unprimed contrast of the present experiment (Figure 8). Notably, the ERP correlate of episodic familiarity (Figure 9) did not include any sign of the early frontal component associated with conceptual priming in the present experiment. 70 22. Discussion
Neural correlates of conceptual priming were characterized using famous faces in a novel
behavioral paradigm. Distinct ERPs were associated with conceptual priming versus explicit memory. Because these two contrasts were derived from the same data, the possibility that observed differences merely reflect confounding task differences can be dismissed. Direct neural
comparisons between conceptual priming and explicit memory, as achieved in the current
experiment, have not been made before. This electrophysiological analysis, taken together with prior ERP findings, thus provided valid insights into the two types of memory.
Priming was observed in every individual tested in the form of faster and more accurate responses to primed than to unprimed famous faces. Perceptual priming was presumably matched between the primed and unprimed conditions, given that each face appeared three times in phase 1. Moreover, possible differences in attentional focus or elaborative processing between primed and unprimed faces would not influence perceptual priming, given the prior demonstration of equivalent perceptual priming under such contrasts (Paller et al., 1999). The priming effects thus belong soundly in the category of conceptual priming.
The frontal positivity found in the priming contrast at 250-500 ms was taken as an ERP correlate of conceptual priming. The effect represents an amplitude reduction in FN400 potentials. In contrast, explicit memory was associated with a posterior positivity at 500-750 ms.
Given that late posterior potentials were also apparent in the primed/unprimed comparison (albeit with smaller amplitudes), one might ask whether the posterior potentials are actually the electrophysiological correlates of conceptual priming. Several arguments suggest not. First, given that the priming manipulation influenced explicit memory, as shown behaviorally in phase
3, a greater degree of explicit retrieval was likely engaged in phase 2 for some primed faces 71 compared to unprimed faces. The primed/unprimed and HEM/LEM contrasts could thus be
expected to share ERP correlates of explicit retrieval (i.e., the late posterior positivity). Second,
reaction-time indices of conceptual priming correlated with early frontal amplitude differences
but not with late posterior amplitude differences. The extent to which the priming manipulation
influenced episodic memory was correlated with late posterior amplitude differences and not
with early frontal amplitude differences. Furthermore, an analysis of the primed/unprimed
contrast restricted to the most well-known celebrities yielded only the early frontal effect. We thus conclude that conceptual priming and explicit memory occurred in conjunction with distinct electrical signals.
The explicit memory test employed did not provide a process-pure measure of familiarity, but rather was meant to index memory for celebrity faces from any source, whether or not the source was also retrieved. If, instead, subjects had been directed to one source only, phase-1 experiences, behavioral measures would have more specifically reflected episodic memory — but this tactic would be problematic because neural measures would likely be contaminated by recall of portions of the extensive pre-experimental knowledge available concerning these celebrities. Fortunately, evidence already available (Figure 9) showed that pure familiarity experiences provoked by repeated faces were associated with late posterior potentials
(Yovel & Paller, 2004). In general, late posterior potentials of the sort elicited in association with explicit memory in the present experiment have been ubiquitously related to episodic memory
(Friedman & Johnson, 2000; Mecklinger, 2000; Paller, 2000; M. D. Rugg & Allan, 2000).
Explicit memory for faces is apparently associated with late, posterior potentials both (a) when retrieval induced by famous faces includes episodic and semantic knowledge (Figure 8), and (b) 72 when retrieval induced by nonfamous faces is restricted so as to support pure-familiarity
experiences (Figure 9).
ERP recordings provide a temporal resolution ideal for examining rapid processing responsible for memory, but they are chiefly sensitive to synchronized postsynaptic potentials generated by neurons situated in a geometric orientation suitable for producing electrical potentials at the scalp. fMRI is subject to different sorts of bias. In fMRI investigations of conceptual priming, frontal and inferior temporal cortex have been implicated, although none of these studies used facial stimuli (Buckner, Koutstaal, Schacter, & Rosen, 2000; Thompson-
Schill, D'Esposito, & Kan, 1999; Wagner, Koutstaal, Maril, Schacter, & Buckner, 2000). One study contrasted brain networks associated with conceptual priming versus explicit memory with words (Donaldson, Petersen, & Buckner, 2001). Because the explicit-memory network did not include as a subset the network associated with conceptual priming, results were used to argue that explicit retrieval did not depend on a contribution from implicit memory. Given that separate tasks were used (abstract/concrete judgments vs. old/new recognition), the pattern of activations could conceivably reflect different task demands per se. Nonetheless, the argument that conceptual priming and explicit memory operate independently is strengthened by the current results, which were not subject to this limitation.
The present results also converge with dissociations observed in amnesia when conceptual priming is spared despite severely impaired explicit memory for learning episodes
(Graf, Shimamura, & Squire, 1985; Keane et al., 1997; Levy et al., 2004; Vaidya et al., 1995).
Here, the neural signature of conceptual priming did not appear to precede ERP correlates of explicit memory for faces (Figure 8) or of pure familiarity for faces (Figure 9). Together, these 73 findings from patients and healthy individuals are consistent with the hypothesis that conceptual
priming and explicit memory rely on distinct neural processes.
The present results also have implications for understanding recollection and familiarity.
Medial temporal structures have been differentially related to recollection and familiarity in
animals (Brown & Aggleton, 2001; Fortin, Wright, & Eichenbaum, 2004), and similar
distinctions have been supported using fMRI in humans. Hippocampal and parahippocampal
activity seems critical for recollection, whereas familiarity is associated with perirhinal activity
(Brewer, Zhao, Desmond, Glover, & Gabrieli, 1998; Davachi, Mitchell, & Wagner, 2003;
Eldridge, Knowlton, Furmanski, Bookheimer, & Engel, 2000; Henson, Cansino, Herron, Robb,
& Rugg, 2003; Ranganath et al., 2004; Yonelinas, Hopfinger, Buonocore, Kroll, & Baynes,
2001). However, fMRI data have not conclusively shown that mutually exclusive neural processes are responsible for recollection and familiarity.
Results presented here provide evidence against the hypothesis that reductions in FN400 potentials reflect familiarity. By disentangling explicit memory and conceptual priming for faces, we showed that reductions in FN400 potentials were strongly associated with conceptual priming. This outcome is in accord with the prior proposal (Olichney et al., 2000) that preserved
N400 reductions with word repetition in amnesic patients reflected spared conceptual priming.
With nonfamous faces, pure familiarity was indexed by posterior positive potentials and
recollection by similar but larger potentials, and neither by reductions in FN400 amplitude
(Yovel & Paller, 2004).
74 Part Five: fMRI Correlates of Familiarity and Conceptual Priming for Famous Faces
23. Rationale
In order to identify relevant brain regions, the experiment described in Part Four was replicated using fMRI to index functional neuroanatomical correlates of familiarity and conceptual priming. In addition to providing spatial specificity to familiarity and conceptual
priming neural measures, employing fMRI measures is also advantageous in that fMRI correlates
of conceptual priming identified using the same novel design used to elicit FN400 correlates of conceptual priming could be compared to fMRI correlates of conceptual priming identified using standard approaches (see below.) In addition, the behavioral paradigm was also modified to provide behavioral and neural measures of familiarity as distinct from other explicit memory processes, whereas the analyses of familiarity in the experiment described in Part Four relied on comparisons with previous results.
24. Methods
Behavioral and fMRI data were obtained from 11 right-handed, native speakers of
English recruited from the Northwestern University community (6 female, ages 20-34), all with
normal or corrected-to-normal vision. Data from an additional four subjects were excluded from all analyses. One excluded subject moved excessively in the scanner and three recognized too few (<80%) celebrity faces during phase 2 of the experiment. 75 Visual stimuli consisted of 180 grayscale images of celebrity faces as well as 180 similar- format images of nonfamous individuals. Three written biographical cues were generated for each celebrity (Appendix 1).
The experiment comprised four distinct phases, represented schematically in Figure 10
(including timing parameters). Phases 1 to 3 mimic those in Experiment 3 of the experiments described above in Part Four (Voss & Paller, 2006), with alterations in stimulus timing made to
accommodate the fMRI environment. Each stimulus was synchronized to the scanner’s repetition
time, and order of conditions was pseudorandomly selected to maximize hemodynamic response
deconvolution.
Phase 1: Biographical matching test. Outside the scanner, subjects were cued to bring to
mind specific conceptual information for primed famous faces but not unprimed famous faces.
Primed famous faces (90) were preceded by a matching biographical cue. Unprimed famous
faces (90) were not preceded by a matching biographical cue, but were instead preceded by an
appropriate description of gender (either “male” or “female”). Nonfamous faces (90) were also
presented, half with a biographical cue matching a primed famous face and the other half with an
incorrect gender description. Subjects responded to each face by pressing either a “match” or
“does not match” button. Phase 1 was divided into three segments (without breaks) with each
face appearing once per segment, each time preceded by a different cue. Faces were shown in
random order. The rapid post-face ISI and response demands during face presentation served to
limit subjects’ ability to incidentally retrieve information pertaining to unprimed famous faces.
The difference in conceptual activation for primed compared to unprimed famous individuals allowed us to obtain measures of conceptual priming in the next phase of the experiment. 76 Phase 2: Conceptual Priming Test. Approximately 15 min after phase 1 (during which subjects were positioned in the MRI scanner), subjects viewed, in pseudorandom order, all 180 famous faces, 60 randomly-selected nonfamous faces from phase 1, and 60 novel nonfamous faces. Subjects responded with a speeded button-press to each famous face and did not respond to nonfamous faces.
We hypothesized that facilitated responses to primed famous faces relative to unprimed famous faces would reflect conceptual priming, given that biographical matching decisions for primed faces in phase 1 selectively engaged access to pertinent conceptual information.
Importantly, the design minimized the influence of other forms of implicit memory on response differences between primed and unprimed famous faces. Perceptual priming was matched for primed and unprimed famous faces, given that each famous face appeared an equal number of times in phase1 with comparable sensory processing. Potential differences in elaborative processing due to task requirements would not be expected to influence perceptual priming, given prior evidence of equivalent face priming regardless of such elaboration (cf. Paller et al.,
1999). In addition, response priming (cf. Dobbins et al., 2004) was equivalent for primed and unprimed famous faces, which were all endorsed with “match” responses during phase 1.
Phase 3: Explicit Memory Assessment. Immediately following phase 2, all faces shown during phase 2 were presented again in a different random order. Subjects rated each face using a four-point scale (Figure 10), indicating the extent to which each face was known. This measure provided an index of the relative amount of nonspecific explicit memory that was likely to have occurred incidentally in response to each face during phase 2.
Phase 4: Episodic Recognition Test. A specific index of familiarity, as distinct from other explicit memory phenomena, was obtained immediately after phase 3. Subjects 77 discriminated 120 nonfamous faces that appeared previously in the experiment (half during all
previous phases, half during only phases 2 and 3) from 30 entirely novel nonfamous faces.
Subjects responded to each face using three buttons corresponding to remember, know, and new.
Remember responses indicated recognition that an item was old accompanied by retrieval of
specific detail from prior episodes, whereas know responses indicated recognition that an item
was old unaccompanied by any such detail. Remember and know responses are commonly used
to index recollection and familiarity, respectively, during recognition testing (Yonelinas, 2002).
During phases 2 and 4, fMRI data were collected using a Siemens TRIO 3.0 T MRI
scanner. Whole-brain gradient-recalled echo-planar images were obtained every 2 s (35 3-mm
axial slices, 0 gap, repetition time=2000 ms; echo time=25 ms; flip angle=80˚; field-of-view=22
cm; 64X64 acquisition matrix; voxel size=3.44X3.44X3 mm, 522 volumes collected during phase 2 and 122 volumes during phase 4). To allow the scanner to reach steady-state, experimental stimuli were not presented during the first 10 volumes (which were discarded).
Following phase 4, high-resolution whole-brain structural images were collected to provide anatomical localization (3D MP-RAGE T1-weighted scans, voxel size=0.859x0.859x1 mm; 160 axial slices).
fMRI analyses were accomplished using the AFNI software package (Cox, 1996).
Preprocessing included coregistration through time to correct brain motion, removal of voxels with low (< 30% of mean whole-brain signal) or erratic (>30% change over one volume) signal, spatial smoothing (7 mm FWHM Gaussian kernel), and transformation to standard stereotactic space (MNI 305). Hemodynamic response deconvolution with a general linear model provided estimates of stimulus-locked neural activity, as quantified using average values from 5-9 s post- stimulus, thus accounting for hemodynamic lag. Regions exhibiting group-level activation 78 differences between experimental conditions were identified via a two-pass random effects
analysis. For each experimental contrast, Monte Carlo simulations performed using the ANFI
program AlphaSim estimated the likelihood of detecting false positives over multiple voxel-wise comparisons. For an individual-voxel probability threshold of p=0.01, we identified the voxel
cluster-size threshold that was necessary to achieve an overall reliability threshold of p=0.01 (cf.
Forman et al., 1995). The most stringent resultant cluster-size threshold, 12 contiguous voxels,
was applied to each contrast.
25. Results
Primed famous faces were identified with greater accuracy than were unprimed famous
faces during phase 2 [mean accuracy for primed = 89.1%, unprimed = 86.4%, t(10)=2.1, p=0.03,
1-tailed], providing a behavioral correlate of conceptual priming. The contrast between primed
and unprimed famous faces identified neural activity within the set of brain regions listed in
Table 2.
To identify neural correlates of nonspecific explicit memory retrieval that occurred
incidentally during phase 2, ratings made during phase 3 were used to classify trials as either
famous faces rated high in explicit memory (HEM; 1 response on the 4-point scale) or as famous
faces rated low in explicit memory (LEM; 2 and 3 responses). Only famous faces correctly
identified as famous in both phase 2 and phase 3 were included. This nonspecific explicit
memory contrast between HEM and LEM famous faces was orthogonal to the conceptual
priming contrast, in that the primed and unprimed famous face categories included roughly equal
numbers of HEM faces on average (52% of HEM faces were primed, SE=5.1%). Neural activity
identified by the nonspecific explicit memory contrast included enhanced positive responses for 79 HEM compared to LEM conditions within the set of brain regions listed in Table 2. These results
were used together with results from the episodic familiarity contrast (below) to obtain a neural
correlate of familiarity with famous faces.
Neural correlates of episodic familiarity for nonfamous faces were identified during
phase 4. Subjects were significantly better than chance at discriminating incidentally-encoded
nonfamous faces from novel nonfamous faces (mean d′=0.93, t(9)=7.9, p<0.001). Familiarity
often occurred in the absence of recollection, as an average of 81% (SE=0.5%) of correct old
trials were know responses. Corresponding neural correlates of were obtained by contrasting fMRI responses to repeated nonfamous faces given know responses versus new nonfamous faces.
This episodic familiarity contrast identified activity within the set of brain regions listed in Table
2.
Neural correlates of conceptual priming and pure episodic familiarity would ideally be obtained during the performance of a single task. However, it is problematic to obtain simultaneous and accurate behavioral estimates of these forms of memory, and this difficulty further confounds attempts to isolate corresponding neural correlates. Relevant to the present circumstances, neural correlates of familiarity-based recognition for famous faces would include other forms of memory that co-occur, such as retrieval of autobiographical episodes or semantic information acquired before the experiment. Thus, comparisons between neural activity identified by the conceptual priming and nonspecific explicit memory contrasts for famous faces would not solely yield neural distinctions between conceptual priming and episodic familiarity, but would identify other divergent forms of memory as well. On the other hand, if results from the conceptual priming contrast for famous faces and the episodic familiarity contrast for 80 nonfamous faces were compared, nonspecific differences in stimuli and task demands would
weaken interpretations.
To overcome these challenges, data from both the nonspecific explicit memory contrast for famous faces and the episodic familiarity contrast for nonfamous faces were pooled in order to identify neural correlates of familiarity for famous faces. These neural measures could then be compared to neural correlates of conceptual priming for famous faces, thus providing neural measures of conceptual priming and familiarity obtained during the performance of a single behavioral task. As outlined above, HEM famous faces were familiar to the subjects, but subjects
were also likely to have recalled specific biographical information about those celebrities
(semantic memories) and perhaps specific episodic information both from phase 1 and from
experiences prior to the experiment. These other explicit memory phenomena may be distinct
from familiarity with respect to neural correlates. Whereas the episodic familiarity contrast and
the nonspecific explicit memory contrast likely identified activity related to many mutually-
exclusive mnemonic phenomena, both contrasts included activity related to familiarity. The
nonspecific explicit memory contrast for famous faces and the episodic familiarity contrast for
nonfamous overlapped in that they both identified a region of significantly enhanced activity in
right inferior parietal lobule (IPL, Figure 11). This right lateral parietal activity is thus taken as a
neural correlate of familiarity within the nonspecific explicit memory contrast.
A double dissociation between conceptual priming and familiarity was evident in fMRI
measures. A double subtraction between the conceptual priming contrast and the explicit
memory contrast indicated that conceptual priming effects were significantly greater in
magnitude than explicit memory effects in regions of left prefrontal cortex (IFG and SFG, Figure
11, Table 2) that were also identified by the main conceptual priming contrast (Table 2). 81 Conversely, neural explicit memory effects were more positive than conceptual priming effects
in the familiarity-related region of IPL (Figure 11). Spatially-overlapping regions were identified
by a double subtraction between the conceptual priming and episodic familiarity contrasts (Table
2). Thus, neural correlates of conceptual priming differed from those of episodic familiarity for
nonfamous faces in the same manner as they differed from neural correlates of explicit memory
for famous faces. Conceptual priming and familiarity were neurally distinct in (1) reliance on
activity within separate anatomical regions, and (2) the polarity of relevant neural processing
(activation reductions for conceptual priming, activation enhancements for familiarity). This double dissociation between conceptual priming and familiarity was evident independent of whether familiarity was indexed generically via the nonspecific explicit memory contrast, or specifically by using results from a recognition memory test with nonfamous faces. Critically, conceptual priming and familiarity were dissociated during the performance of a single task because conceptual priming differed from nonspecific explicit memory in the region of IPL that was associated with familiarity within the nonspecific explicit memory contrast.
26. Discussion
Neural measures of conceptual priming and familiarity were both elicited by famous
faces during the performance of a conceptual priming task. By making use of multiple behavioral
measures in a novel paradigm, neural indices of these distinct memory events were dissociated.
This finding constitutes the first dissociation between neuroimaging measures of familiarity and conceptual priming within fMRI data obtained during the performance of a single behavioral
task. 82 Conceptual priming for famous faces was associated with response reductions in two regions of left prefrontal cortex, IFG and SFG. Response reductions in similar prefrontal regions have been identified in other conceptual priming tests with verbal materials (Donaldson et al.,
2001; e.g. Thompson-Schill et al., 1999; Wagner et al., 2000), supporting the validity of the novel method used here to induce conceptual priming. Conceptual priming with both accuracy and response-time measures was also observed in three previous experiments using similar procedures [Part 4; identical task requirements in Experiment 3 of Voss & Paller (2006); slightly modified procedures in Experiments 1-2 of Voss & Paller (2006)], and yet, priming effects in the present experiment were not apparent in reaction times and were smaller in accuracy measures.
As priming effects generally decline sharply as retention delay increases (Gabrieli, 1998;
Schacter & Buckner, 1998), we attribute the reduced magnitude of priming here to the delay associated with moving subjects into the fMRI scanner prior to Phase 2. Nevertheless, the design was successful for revealing reliable neural conceptual priming effects.
To identify neural correlates of familiarity for famous faces, we pooled information across contrasts that individually identified neural correlates of nonspecific explicit memory for famous faces and familiarity for nonfamous faces. Because multiple forms of memory are mobilized in response to repeated stimuli irrespective of whether corresponding behavioral measures are also provided, it is not possible to administer “process-pure” tests of familiarity for famous faces. In other words, if a recognition test was administered to assess memory for phase
1 episodes with famous faces, neural measures of familiarity would likely be contaminated by activity related to semantic retrieval of various facts known about each celebrity, even if such retrieval was not required of subjects. Behavioral measures would be insufficient to rule out incidental retrieval of this sort in order to disambiguate the pattern of fMRI results. We thus 83 utilized a nonspecific behavioral measure of explicit memory that was potentially sensitive to a
variety of mnemonic phenomena that could have influenced neural measures. Fortunately, we
were able to identify neural correlates of familiarity-based recognition memory for nonfamous faces in phase 4, and we used this information to identify the components of retrieval associated with familiarity-based recognition in phase 2. Using this method, response enhancements in right
IPL were attributed to familiarity for famous faces. Many prior studies using verbal materials have associated left IPL activation enhancements with recognition memory (reviewed in
Wagner, Shannon, Kahn, & Buckner, 2005) and, specifically, familiarity-based recognition
(Wheeler & Buckner, 2004). A tenable interpretation is that familiarity-related activity is mirrored onto the contralateral hemisphere in the present study given that the right-hemisphere plays a dominant role in face processing and that right parietal lobe activity has been found in association with face memory (Sperling et al., 2001).
The paradigm employed here to elicit neural correlates of explicit and implicit memory phenomena involved the use of a variety of behavioral measures in order to substantiate relationships between memory phenomena and corresponding neural measures. Whereas these results are in harmony with the general distinction that has been made in the fMRI literature between the polarity of relevant neural processing for explicit memory (response enhancements) and implicit memory (response reductions), this does not preclude the possibility that this pattern might not persist under different circumstances. For instance, familiarity has been associated with response reductions in entorhinal cortex (Henson et al., 2003) and heightened responses have been associated with implicit memory (Henson, 2003). It is critical that any neuroimaging examination of memory include appropriate behavioral measures of potentially-related phenomena in order to accurately attribute memory events to underlying brain activity. 84 In a previous fMRI study (Wheeler & Buckner, 2004) neural responses associated with
conceptual priming were modulated exclusively during a lexical decision (implicit) test, whereas
responses associated with recognition memory were modulated during both the implicit test and
a recognition memory (explicit) test. Because implicit test activity was not evident during the
explicit test, the authors concluded that recognition performance was not driven by conceptual
priming. Instead, it is possible that conceptual priming-related neural processing varies with task
demands. Thus, conceptual priming activity that was not present during the implicit test could have occurred during the explicit test, yet been unidentified. The present design was not subject
to the same limitations, and the observed dissociation can thus be taken as direct support for the
notion that conceptual priming and familiarity are functionally independent.
Similar procedures (described in Part 4) were used to elicit ERP correlates of conceptual
priming and explicit memory for famous faces (Voss & Paller, 2006). Spatio-temporally distinct
ERPs were identified for these memory phenomena, with early-onset ERPs maximal over the
front of the head attributed to conceptual priming and later-onset ERPs with a posterior
distribution attributed to explicit memory. This dissociation in ERP activity is entirely consistent
with the present findings. The present results extend the ERP findings by including a specific
measure of familiarity, whereas ERP analyses of familiarity had been derived from data collected
in another study (Yovel & Paller, 2004). Furthermore, the present fMRI results add anatomical
specificity that was not provided by scalp-recorded ERP data.
85 Part Six: ERP Correlates of Familiarity and Conceptual Priming for Minimalist Visual Shapes
27. Rationale
FN400 and left-frontal fMRI correlates of conceptual priming were obtained during a test of conceptual priming in Part Four and Part Five, whereas nearly all of the ERP familiarity literature concerns FN400 potentials recorded during recognition testing. It was thus necessary to determine if FN400 potentials obtained during recognition can also reflect conceptual priming.
This is difficult, however, in that it is problematic to measure conceptual priming during a recognition test.
To overcome this challenge, minimalist visual shapes that conveyed minimal meaning were employed as stimuli during recognition testing. These squiggles varied in the degree to which they resembled meaningful objects, and the perceived meaningfulness for any given squiggle varied idiosyncratically across viewers. Because conceptual priming could occur only for the most meaningful squiggles whereas familiarity-based recognition could occur largely independent of meaningfulness, it was thus possible to identify ERP correlates of both familiarity and conceptual priming by tracking meaningfulness and familiarity in each subject. This allowed for a comparison of neural correlates of both memory phenomena using neuroimaging measures obtained during recognition testing.
28. Methods
Visual stimuli consisted of 300 squiggles (Figure 13). Squiggles were presented on a computer monitor in black on a white background within a square subtending approximately 5° 86 of visual angle. Squiggles were taken from a recent study of explicit memory (Groh-Bordin et al., 2006) and were created via hand-deformation of a square, circle, or triangle.
In Experiment 1, behavioral data were collected from 12 right-handed native English speakers (4 males, ages 18-23 yrs) recruited from the Northwestern University community. The experiment consisted of nine study-test blocks during which subjects viewed all 270 squiggles.
Blocks were separated by a short break. Each block consisted of a study phase followed by one of three possible tests: loop discrimination, implicit rating, and explicit rating. Each test was administered three times in randomized order, and subjects were unaware of the total number of blocks such that test format could not be determined during the study phases.
Study Phase In each block subjects viewed 20 squiggles, each for 2000 ms with a variable 1500-3000 ms interstimulus interval (ISI). Subjects rated each squiggle using the 4- point meaningfulness scale with 1 corresponding to “high meaningfulness,” and 4 to “no meaningfulness.” Subjects were instructed to rate squiggles as 1 if the squiggle “looks like a nameable object, face, or animal” and as 2 if the squiggle “looks like a more abstract nameable object, face, or animal.” A rating of 3 indicated that the squiggle “does not look like anything nameable, but is in some way meaningful.” Subjects were provided the example that the squiggle
“may be angled such that it appears to be angry.” A rating of 4 corresponded to “a random collection of lines that is in no way meaningful.” Subjects were instructed to distribute ratings across the four levels. We operationally defined squiggles given meaningfulness ratings of 1 or 2 as high in subjective meaningfulness (High-M) and those given ratings of 3 or 4 as low in subjective meaningfulness (Low-M). Subjects were made aware that memory for squiggles would be tested subsequently, and that test format would vary randomly. 87 Test Phase The test phase followed the study phase in each block after a break of
approximately 45 s during which subjects counted backwards aloud by threes from a designated
integer for 20 s and then were given test-phase instructions. Each test consisted of 20 squiggles
repeated from the previous study phase (old) and 10 entirely novel squiggles (new), presented in
randomized order, each for 1000 ms with a variable 1000-2000 ms ISI. The three possible test formats are described below.
Meaningfulness Rating Test To index conceptual priming for squiggles, subjects rated the meaningfulness of each squiggle using the 4-point scale described above. Response speed was emphasized. Subjects were told that they had seen some of the stimuli previously, and that they should disregard this prior exposure as well as the rating made previously because attending to this information could slow responses. Responses were made using the right hand, with two assignments of meaningfulness rating to response finger counterbalanced across subjects.
Adjacent fingers corresponded to adjacent ratings with either 1 pressed by the index finger and 4 by the little finger or vice versa.
Meaningfulness Rating Test with Repetition Acknowledged To determine if conceptual
priming effects observed during the implicit rating test could be due to explicit remembering of
study phase stimuli or ratings, the implicit rating test was performed with the modification that
subjects also make a mental note of whether each stimulus was old or new. Following each test,
subjects provided a rough estimate of the approximate number of repeated and novel stimuli, and
were not provided feedback. Rating response speed was emphasized, and subjects were instructed to refrain from keeping a mental tally of old or new stimuli, as this would slow them down. 88 Loop Discrimination Test To index perceptual priming for squiggles, subjects indicated
the presence or absence of a loop in the stimulus by pressing one of two buttons (50% of stimuli contained a loop, see Figure 13). Response speed was emphasized.
In Experiment 2, behavioral data were collected from 10 right-handed native English speakers (4 males, ages 18-21 yrs) recruited from the Northwestern University community.
A set of 150 squiggles (selected randomly for each subject from the total set) were shown for
1000 ms with a variable ISI of 1500-2500 ms. All squiggles were presented for a second time at a variable delay of 5 to 15 trials (average delay of 20 s) from initial presentation. A randomly selected 75 squiggles were presented for a third time at a variable delay of 20 to 30 trials
(average delay of 50 s) from initial presentation.
Subjects rated squiggles using the 4-point meaningfulness scale described above.
Subjects were informed that squiggles would repeat and were advised to make each rating irrespective of previous ratings. Response speed was emphasized. Responses were made using the right hand, with two assignments of meaningfulness rating to response finger counterbalanced across subjects. Adjacent fingers corresponded to adjacent ratings with either 1 pressed by the index finger and 4 by the little finger or vice versa.
In Experiment 3, behavioral and ERP data were collected from 15 right-handed native
English speakers (7 males, ages 18-35 yrs) recruited from the Northwestern University community. Three sets of 100 squiggles were created via random assignment for counterbalancing, such that each squiggle appeared as a new one for five subjects and as an old one for all other subjects. The experiment consisted of ten study-test blocks during which subjects viewed all 300 squiggles. Blocks were separated by a short break. Prior to experimental blocks, subjects completed an abbreviated practice block using an addition set of stimuli that 89 were not included in the main experiment. Verbal instructions preceded every study and test
phase throughout the experiment. Blocks were identical to blocks in Experiment 1 in stimulus timing parameters and study-test structure with one exception – each test phase was a recognition
test instead of one of three priming tests. After an average delay of ten months (range 8-11
months), ratings were collected for all squiggles such that new test-phase stimuli could be
divided into meaningfulness categories. Ratings to old stimuli were highly consistent with those
provided initially (91% of squiggles were assigned to the same meaningfulness category,
SE=2.3%).
ERP Test Phase Subjects used four buttons to categorize each squiggle as old or new,
with four response categories based on a modified “remember/know” paradigm (Gardiner &
Java, 1991; Tulving, 1985). The categories (shown in Figure 14) were: (1) high-confidence
recollection of specific study-phase episodic detail (remember responses), (2) high-confidence
recognition unsubstantiated by specific detail (know responses), (3) low-confidence guess
responses, or (4) indication that the stimulus did not appear during the study-phase. The practice
phase was used to ensure that subjects adopted appropriate criteria for each response category.
Continuous electroencephalographic recordings were made during study and test phases
from 59 evenly distributed scalp sites (Woldorff et al., 2002) using tin electrodes embedded in an
elastic cap. Four additional channels were used for monitoring horizontal and vertical eye
movements, and only artifact-free trials were included in ERP analyses (average of 89% of trials
per subject, SE=0.06%). Electrode impedance was ≤5 kΩ. EEG signals were amplified with a
band pass of 0.05 to 200 Hz, sampled at a rate of 1000 Hz, and rereferenced offline to average
mastoids. Each averaging epoch lasted 1100 ms, including 100 ms prior to stimulus onset.
Baseline correction was performed by subtracting prestimulus mean amplitudes. 90 Analyses focused on test-phase electroencephalographic responses. ERPs elicited by squiggles during the test phase were averaged separately for each response type (remember, know, guess, and new) and as a function of meaningfulness ratings made to old items when they had appeared in the study phase (High-M and Low-M). Trials were included in analyses if a
correct response was given in the test phase. Statistical comparisons focused on amplitudes
averaged over anterior, middle, and posterior regions (Figure 15a). Visual inspection of ERPs
from individual electrodes confirmed that spatially averaged data from the three scalp regions
adequately characterize the experimental effects.
Formal comparisons of ERP amplitude were made using repeated-measures ANOVA
(α=0.05) with Huynh-Feldt corrections when necessary. Post-hoc pairwise comparisons were made between conditions for each region and latency interval and type-I error was controlled via
Bonferroni correction. Waveforms were smoothed with a 10-Hz low-pass-zero-phase-shift
Butterworth filter for presentation purposes only.
29. Results
An analysis of response times collected during Experiment 1 confirmed that conceptual priming occurred preferentially for relatively meaningful squiggles (Table 3). The implicit memory test required a rating of squiggle meaningfulness, and a measure of conceptual priming was obtained by comparing response times for repeated versus new squiggles. Responses were speeded by 48 ms for repeated squiggles that were initially given high meaningfulness ratings
(High-M, 43% of stimuli on average, SE=2.5) but there was no speed-up for those given low meaningfulness ratings (Low-M). 91 Furthermore, this conceptual priming was disrupted when subjects were instructed to focus some attention on stimulus repetition. In this case, there was no response speed-up for either High-M or Low-M squiggles.
In contrast, equivalent perceptual priming was found for these categories using an implicit memory test that required squiggles to be discriminated on the basis of whether a loop was present. Discrimination was more accurate for both High-M and Low-M compared to new squiggles. Magnitude of priming (mean 5.3% accuracy improvement) did not differ significantly for High-M versus Low-M squiggles (p=0.73).
Although the High-M/Low-M contrast is akin to a systematic manipulation of depth of processing, squiggles were categorized by subject ratings instead of counterbalanced assignment.
Nonetheless, a high degree of rating variability for individual stimuli produced an intrinsic counterbalancing. Each squiggle was just about as likely to be assigned to one of the meaningfulness categories as to the other (mean 47% chance of falling into the High-M category,
SE=3%). Priming effects thus cannot be readily attributed to nonspecific differences in stimuli comprising the two meaningfulness categories.
Similar conceptual priming effects were identified using a continuous presentation format in Experiment 2. Ratings were made increasingly faster with repetition for squiggles given high meaningfulness ratings (High-M) but not for those given low ratings (Low-M). On average, 44%
(SE=0.04) of squiggles were given High-M ratings. Mean response time across the three repetitions was 949, 891, and 858 ms, respectively, for High-M ratings and 919, 920, and 923 ms for Low-M. Repeated-measures ANOVA with two factors, condition (High-M/Low-M) and repetition (first/second/third presentation), yielded a significant main effect of repetition
[F(2,18)=4.5, p=0.03] and a significant interaction [F(1.9,17.3)=6.17, p=0.01]. High-M response 92 times decreased significantly with repetition [first vs. second t(9)=2.4, p=0.04; second vs. third t(9)=4.8, p=0.001] whereas Low-M response times did not [first vs. second t(9)=0.06; second vs. third t(9)=0.12].
The possibility that response priming (cf. Dobbins et al., 2004) contributed to conceptual priming effects is unlikely given results from subsets of stimuli rated inconsistently — stimuli that received either inconsistent High-M (1 or 2) or inconsistent Low-M (3 or 4) ratings across the first two presentations. Behavioral priming was observed for the inconsistently rated High-M squiggles [average decrease = 42 ms; t(7)=4.3, p=0.004], but not for the inconsistently rated
Low-M squiggles [average increase = 23 ms; t(7)=1.4, p=0.22]. Thus, response facilitation for
High-M squiggles was not due merely to strengthened stimulus-response mapping.
The behavioral response speed-up attributed to conceptual priming in both Experiment 1 and Experiment 2 was not due merely to an influence of meaningfulness on rating speed, but rather reflected repetition-induced facilitated access to stimulus meaning. When squiggles were rated during encoding in Experiment 1, response times to High-M and Low-M stimuli were equivalent (1375 ms and 1364 ms on average, respectively, p=0.64). Likewise, response times to
High-M and Low-M stimuli were equivalent for the first presentation of each stimulus during
Experiment 2 (949 ms and 919 ms on average, respectively, p=0.49).
These behavioral results taken together demonstrate that the response facilitation for
High-M squiggles cannot be readily attributed to perceptual priming, to explicit remembering of the initial encounter, to response priming, or to any confounding effects of stimulus factors. The priming effect for High-M squiggles thus belongs in the category of conceptual priming.
During Experiment 3, ERPs were recorded in a study-test structure identical to that used to assess conceptual priming in Experiment 1, except that recognition was assessed in the test 93 phase. “Remember” and “know” responses were used by subjects to indicate phenomenological features of episodic memory retrieval, recollection and familiarity, respectively (Yonelinas,
2002). Although the validity of this approach has been questioned (Wixted, in press), here it is only assumed that remember and know responses that subjects make provide a good approximation to corresponding self-rated subjective experiences of recollection and familiarity
(not hypothetical processes also known as recollection and familiarity).
Recognition sensitivity (d') was calculated separately for each response type to determine the extent to which the behavioral responses used to categorize ERPs reflected veridical memory.
Subjects successfully distinguished old from new squiggles (Figure 14a) using both remember and know judgments (mean remember d'=2.6, SE=0.43, mean know d'=0.7,SE=0.14). The frequency of guess responses, however, was similar for old and new squiggles (mean guess d'=
-0.4, SE=0.04).
To assess the influence of stimulus meaningfulness on memory, study-phase ratings were used to divide old squiggles into High-M and Low-M categories of approximately equal numbers
(Figure 14b). An analysis of test performance calculated separately for these categories (Figure
14c) indicated superior memory for High-M than Low-M squiggles. This improvement was due to more remember responses [t(14)=9.6, p<0.001] and fewer guess and new responses
[t(14)=8.7, p<0.001 and t(14)=7.5, p<0.001, respectively]. In contrast, the number of know responses was approximately equivalent for High-M and Low-M [t(14)=0.43]. These findings show that higher meaningfulness led to stronger recollection. However, one might question whether familiarity per se was equivalent for High-M and Low-M squiggles, given that familiarity also likely occurred for trials categorized by remember responses. However, if we take know responses as an indication of the behavioral phenomenon of pure familiarity without 94 recollection (as opposed to a hypothetical cognitive process), then these results imply that the
High-M and Low-M conditions produced a very similar number of trials that engendered the
experience of familiarity without recollection.
An analysis of study-phase rating variability highlighted the importance of assessing
meaningfulness separately for each subject. Although some squiggles were endorsed by the
majority of subjects as either High-M or Low-M (Figure 13a), most squiggles were not given
consistent ratings (Figure 13b). Typically, squiggles were rated as High-M and Low-M,
respectively, by approximately equal numbers of participants (Figure 14d). Thus, relying on
normative meaningfulness ratings (Groh-Bordin et al., 2006) rather than assessing stimulus
meaning on an individual basis would not accurately characterize effects of inferred meaning on
implicit memory or on neural correlates of memory.
In order to facilitate comparison with prior studies that did not employ meaningfulness
ratings or remember/know judgments, we first identified neural correlates of explicit memory by
averaging ERPs to all correctly endorsed old and new squiggles (Figure 15b). ERPs to old
squiggles were more positive compared to ERPs to new squiggles over most of the scalp starting
at approximately 300 ms post-stimulus. Inspection of scalp topographic maps (Figure 15c)
suggested that ERP differences consisted of an early component, centered over the front of the
head and beginning at approximately 300 ms, and a late component, centered over the rear of the
head and beginning at approximately 400 ms.
Statistical comparisons made between ERPs averaged over three regions and three
latency intervals (Figure 15a) yielded a significant repetition effect [F(1,14)=19.6, p<0.001] and
3-way interaction [F(1.9,27.2)=3.51, p<0.05]. Old ERPs were reliably more positive than new 95 ERPs in the anterior region from 300-500 ms and in the middle and posterior regions from 500-
700 ms and 700-900 ms (all p’s<0.006).
The overall old/new repetition effect described above was fractionated according to the
putative behavioral indices of recollection and familiarity (i.e., remember and know responses, respectively). Initially, this analysis was conducted independent of stimulus meaningfulness.
Compared to new squiggles, old squiggles in remember or know categories occurred with more
positive ERPs beginning at approximately 250 ms, with overall greater amplitudes for remember
(Figure 16). Scalp topographic maps of the remember/new difference consisted of an early onset
(approximately 300 ms) frontal-maximum positive difference and a late onset (approximately
500 ms) centroparietal-maximum positive difference, similar to the overall old/new difference.
Know/new ERP differences also included a late-onset centroparietal-maximum positivity, but at
early latencies differences were relatively diffuse on the scalp.
Formal comparisons of these ERP differences were conducted over three regions and
over the latency intervals from 300-500 ms and 500-700 ms in order to distinguish early from
late ERP effects. ERPs for both remember and know differed from new ERPs across regions and
latency intervals [3-way interactions: F(1.3,17.5)=14.6, p<0.001; F(1.2,16.4)=2.86, p=0.08, respectively]. From 300-500 ms, compared to new ERPs, know ERPs were more positive in the anterior region and remember ERPs were more positive in the anterior and middle regions; from
500-700 ms, both remember and know ERPs were more positive than new ERPs in the posterior region (all p’s<0.008). Differences between remember and know ERPs depended on latency interval and region [3-way interaction: F(1.1,14.2)=6.5, p=0.02], with greater amplitudes for remember during the later interval in the posterior region [t(14)=5.6, p<0.001], but not in other regions or latency intervals. 96 Meaningfulness ratings were used to identify neural correlates of conceptual priming in contradistinction to those of explicit memory. This was accomplished by averaging ERPs to squiggles given know responses separately for High-M (conceptual priming present) and Low-M
(negligible conceptual priming) categories. Because know responses were given with identical response criteria and in similar numbers for both High-M and Low-M squiggles, this contrast identified ERP correlates of conceptual priming with explicit memory held constant. Four subjects were excluded from this analysis due to an inadequate number of High-M or Low-M trials (<20).
The ERP difference associated with conceptual priming (Figure 17) was apparent over frontal electrodes at all latencies after approximately 200 ms and over posterior electrodes starting at approximately 400 ms. Based on our a priori hypotheses, mean ERP amplitude for
High-M and Low-M know responses was compared over anterior, middle, and posterior regions for the 300-500 and 500-700 ms intervals. The conceptual priming difference was reliable in the anterior region during both intervals (p’s<0.03) but not in middle and posterior regions during either interval (p’s>0.10).
The validity of this conceptual priming contrast depends on equivalent explicit memory strength for High-M-Know and Low-M-Know squiggles. Clearly, overall explicit memory was stronger for High-M compared to Low-M squiggles. If the conceptual-priming contrast had included remember responses, then neural correlates of stronger explicit memory would be confounded with those of stronger conceptual implicit memory. The contrast thus focused on a subset of squiggles, those given know responses, which were equivalent across meaningfulness levels. 97 It is possible that ERP correlates of conceptual priming reflected an influence of inferred
meaning on ERPs rather than an effect of memory per se. To assess this possibility, ratings
collected following ERP recordings were used to categorize new squiggles viewed during the
test phase into High-M and Low-M categories. There were no systematic amplitude differences
over the three regions and two intervals (all p’s > 0.23), and no striking differences were
observed at any electrode for any latency. Thus, ERP correlates of conceptual priming were
indeed repetition-related memory effects.
A close connection between familiarity and late posterior potentials (and not between
familiarity and early anterior potentials) was supported by across-subject correlations between behavioral and ERP measures. To index memory strength for pure familiarity responses, d' was calculated separately for High-M and Low-M squiggles recognized with know responses (using false-alarm rates derived from know responses to new stimuli). ERP differences were quantified for each region and latency interval for the contrast between squiggles recognized with know responses and new squiggles. In order to account for individual differences in the spatial focus of each ERP effect, a single electrode was selected in each case according to where the greatest between-condition difference was observed. For both High-M and Low-M squiggles, know response d' was significantly correlated with know-versus-new ERP amplitude differences from
500-700 ms at middle [High-M r(9)=0.62, p=0.04; Low-M r(9)=0.69, p=0.02] and posterior
[High-M r(9)=0.64, p=0.03; Low-M r(9)=0.79, p=0.004] regions (Figure 18). No correlations involving the anterior region or the early latency interval reached statistical significance
(p’s>.20). 98 30. Discussion
Using behavioral measures of multiple types of memory, we identified ERP correlates of
explicit memory and of conceptual implicit memory during a recognition test. Remember/know judgments were used to index the phenomenological experiences of explicit recognition known in the literature as recollection and familiarity. When data were analyzed irrespective of any consideration of conceptual priming, as is a common practice in contemporary studies of neural correlates of recognition, two types of ERPs were associated with recognition, late-onset posterior maximum positive potentials (a “parietal old/new effect”) and early-onset frontal- maximum positive potentials (a “FN400 old/new effect”).
The present experiment provided the first evidence to link specific ERPs to conceptual priming for visual objects. Based on our prior experiments with other visual stimuli (Olichney et al., 2000; Voss & Paller, 2006; Yovel & Paller, 2004), we predicted that FN400 old/new effects would be associated with conceptual priming. We attributed FN400 effects observed here
(Figure 17) as correlates of conceptual implicit memory based on the following reasoning. First, behavioral results demonstrated conceptual priming selectively for the stimuli of highest perceived meaning. Although these squiggles were not very meaningful, subjects nevertheless were able to envision, on an idiosyncratic basis, some resemblance to meaningful visual objects.
We excluded other possible explanations for the speed-up in responses to repeated squiggles, such that conceptual priming could be inferred. Second, we compared neural responses across conditions that differed selectively in conceptual priming. In an analysis of trials wherein old squiggles were recognized with pure familiarity, we contrasted ERPs to squiggles capable of engendering conceptual priming versus ERPs to squiggles subject to limited or no conceptual priming. This contrast thus constituted a manipulation of conceptual priming with explicit 99 memory held constant.
Whereas FN400 ERPs were associated with conceptual priming, later parietal ERPs were
associated with familiarity-based recognition. This association was marked first by virtue of the
old/new ERP contrast restricted to old squiggles that gave rise to behavioral responses signaling
pure familiarity (Figure 16). This relationship was substantiated by a correlational analysis run
across subjects, whereby a behavioral measure of familiarity sensitivity was systematically and
selectively related to the magnitude of late parietal old/new ERP differences (Figure 17). As
opposed to what might have been predicted based on the hypothesis that FN400 ERPs index
familiarity (Curran, Tepe et al., 2006), the magnitude of behavioral familiarity was not correlated
with FN400 ERP differences. In sum, ERPs elicited during this recognition test were found to
reflect both explicit memory and conceptual priming, in the form of parietal old/new effects and
FN400 old/new effects, respectively.
These results have important implications for understanding memory functions and for
interpreting neuroimaging evidence obtained using recognition tests. The identification of distinct neural correlates of episodic recollection in contradistinction to episodic familiarity has been taken as strong empirical support for dual-process accounts of recognition memory
(Yonelinas, 2002). Dual-process models assert that qualitatively distinct recollection and familiarity processes support recognition memory. In ERP experiments, recollection has been ubiquitously linked to parietal old/new effects (Friedman & Johnson, 2000; Mecklinger, 2000;
Paller, 2000; M. D. Rugg & Allan, 2000). A large set of experimental reports have attributed frontal N400 old/new effects to familiarity (recently reviewed by Curran, Tepe et al., 2006).
However, an alternate interpretation supported by the present and other findings (Olichney et al.,
2000; Voss & Paller, 2006; Yovel & Paller, 2004) is that FN400 old/new effects instead reflect 100 conceptual implicit memory elicited incidentally during episodic memory testing (for review, see
Part Three and Paller, Voss, & Boehm, in preparation). Although the published evidence on this point is mixed, discrepancies concerning these issues can be partly attributed to the failure of studies that employed meaningful stimuli to adequately control for or measure conceptual priming. Given that interpretations of many neuroimaging results have failed to take conceptual priming into account, support from these studies for dual-process theories, and for hypotheses about the neural substrates of familiarity, can thus be called into question.
Concern about adequately accounting for conceptual priming in studies of recognition is not limited to ERP research. Notably, fMRI has provided neuroanatomical support for dual- process models of recognition memory in that recollection and familiarity occur with activity in segregated brain regions (Davachi et al., 2003; Eldridge et al., 2000; Henson et al., 2003;
Ranganath et al., 2004; Yonelinas et al., 2001; Yonelinas et al., 2005). Neuroanatomical correlates of recognition have been found not to include activity related to conceptual priming as identified during specialized implicit memory tests (Donaldson et al., 2001). There is scant evidence, however, to show whether neural correlates of conceptual implicit memory differ when elicited during implicit memory testing versus during episodic memory testing. Future fMRI comparisons should address this concern by contrasting functional neuroanatomical correlates of conceptual priming and explicit memory identified during the performance of a single task (Part
Five).
The results presented here provide a foundation for critically assessing the hypothesis that episodic familiarity is driven by conceptual implicit memory (Rajaram & Geraci, 2000;
Verfaellie & Cermak, 1999; Wagner et al., 1997; Wolk et al., 2005). Evidence supporting this hypothesis would be obtained if neural correlates of conceptual implicit memory were generally 101 found to precede those of episodic familiarity, and if both reliably occurred together during recognition. Effective connectivity analyses might also provide relevant evidence. However, compelling dissociations between conceptual priming and explicit memory have been provided by neuropsychological studies in amnesic patients (Graf et al., 1985; Keane et al., 1997; Levy et al., 2004; Vaidya et al., 1995). Research that might verify such dissociations in healthy individuals by directly examining putative functional relationships can now be pursued by making use of neural correlates of memory as characterized in the present study.
102 Part Seven: ERP Correlates of Familiarity and Conceptual Priming for Words
31. Rationale
During recognition testing, ERP correlates of conceptual priming for squiggles were
found to include FN400 potentials, whereas familiarity-based recognition was found to produce
later-onset parietal-centered potentials. However, minimalist visual shapes are seldom used as
stimuli for recognition testing. It was thus essential to determine if conceptual priming during
recognition for standard stimuli, such as words, can include FN400 potentials. However, it is
problematic to obtain neural correlates of conceptual priming for common words because
priming occurs to all repeated common words to a similar extent.
To produce variations in conceptual priming for words, esoteric words were used as
stimuli for recognition testing, using a similar paradigm as described in Part 6. Subjects were misled to believe that these uncommon words were not words at all, and rated the abstract meaningfulness of each word in order to provide a behavioral estimate of conceptual priming
(present for only the most meaningful items) that was largely independent from familiarity-based recognition (present to a similar extent irrespective of meaningfulness). This allowed for ERP correlates of familiarity and conceptual priming to be obtained for verbal materials during recognition testing.
32. Methods
Eighteen right-handed, native speakers of English provided behavioral and electrophysiological data (six male, ages 19-26). Data from an additional five subjects were 103 excluded: three due to excessive eye-blink artifacts and two because of the exclusive use of
remember responses during recognition testing (see below).
Visual stimuli were 240 extremely uncommon English words (Appendix 2) between 4
and 10 letters in length. Words were presented in Arial font in black on a white background at
central fixation, and subtended approximate visual angles of 0.5˚ vertically and between 1.5˚ and
3.5˚ horizontally. A central fixation cross was present during interstimulus intervals.
There were eight experimental blocks, each consisting of a study phase and a test phase.
During each study phase, 20 words were presented individually (2000 ms presentation, 2000-
3500 ms randomized ISI). Subjects were misinformed that each stimulus was either an uncommon word or a pseudoword, and rated each stimulus for meaningfulness using a 5-point scale where 1 indicated that the item was a real word, and 2 through 5 corresponded to high-, medium-, low-, and negligible-meaningfulness, respectively, for stimuli identified as pseudowords. Instructions were to rate a stimulus as high if it immediately invoked a concrete meaning, medium if it immediately invoked an intangible meaning or connotation, low if it was possible to attribute minimal meaning to the stimulus with effort, and negligible if the stimulus invoked no meaning. Items endorsed as English words (9.8% on average, SE=0.8%) were excluded from all behavioral and ERP analyses. Subjects were instructed to distribute ratings across meaningfulness categories and were made aware that memory would be subsequently tested via a practice block utilizing an additional set of uncommon words that was completed before the main experiment.
Recognition memory was tested following each study phase after a delay of approximately one minute during which subjects were given instructions. Each test consisted of the 20 words viewed during the previous study phase and 10 entirely novel words, presented in 104 randomized order (1000 ms presentation, 2000-3000 ms randomized ISI). Novel words (80 total) were counterbalanced across subjects. Subjects used five buttons to categorize each stimulus as old or new using response categories based on a modified “remember/know” paradigm (Gardiner
& Java, 1991; Tulving, 1985), intended to index recollection and familiarity strength. Subjects responded remember when specific study-phase episodic detail was recollected with high confidence. Know responses indicated recognition without retrieval of detail, and were subdivided into high-, medium-, and low-confidence categories. Subjects responded new to indicate that the stimulus did not appear during the study phase.
Scalp electroencephalographic recordings during the study and test phases were made from 59 evenly distributed tin electrodes embedded in an elastic cap (Woldorff et al., 2002).
Voltage was referenced to a right mastoid electrode, and rereferenced offline to average mastoids. The electrooculogram was recorded from four additional channels (below the center of each eye and on each outer canthus). Electrode impedance was kept below 5 kΩ. EEG signals were recorded with a band pass of 0.05 to 200 Hz, and sampled at a rate of 1000 Hz. Each 1100 ms averaging epoch began 100 ms prior to stimulus onset. Mean prestimulus amplitudes were subtracted to correct for baseline variability. Epochs containing EEG artifacts were excluded from ERP analyses (average of 8.8% per subject, SE=1.4%).
ERPs were averaged selectively as a function of test phase responses and study phase meaningfulness ratings. Words given high and medium meaningfulness ratings were operationally defined as high in subjective meaningfulness (meaningful), and those given low and negligible ratings as low in subjective meaningfulness (meaningless). Statistical comparisons were performed on amplitudes averaged over seven scalp regions (Figure 19), and were made 105 using repeated-measures ANOVA (α=0.05) with Huynh-Feldt correction for nonsphericity. Post-
hoc pairwise comparisons utilized Bonferroni correction for multiple comparisons.
33. Results
Subjects categorized their recognition experience using remember and know responses
(to index the phenomenological experiences of recollection and familiarity, respectively)
differently for meaningful and meaningless words [meaningfulness-by-response interaction
F(4,68)=25.8, p<0.001]. Overall stronger recognition for meaningful compared to meaningless
words [main effect of meaningfulness F(1,68)=8.9, p<0.05] was due to a significantly greater
number of remember responses, fewer low-confidence know responses, and fewer new
responses. In contrast, the number of high- and medium-confidence know responses did not
differ significantly between meaningful and meaningless words (Table 4).
To determine the extent to which subjects’ use of each recognition response category
reflected veridical memory, discrimination sensitivity (d′) was calculated separately for each
response category as a function of meaningfulness category (Table 4). Discrimination of old
from new items was successful when subjects used both remember and high-confidence know response categories, as indicated by positive d′ values that were significantly greater than zero. In
contrast, discrimination failed for medium- and low-confidence know responses, for which d′ was less than or not significantly different from zero.
To identify neural correlates of conceptual priming using the present paradigm, it was necessary to contrast ERP responses to two categories of words that differed in meaningfulness yet were matched in explicit memory strength. Meaningful words recognized with high- confidence know responses and meaningless words also recognized with high-confidence know 106 responses were two categories sufficient for this purpose. Theoretically, words given lower
meaningfulness ratings have less inherent meaning than those given higher ratings. This
assertion was substantiated by subjects’ superior recognition performance for meaningful
compared to meaningless words, as recognition is enhanced as a function of meaningfulness
(Curran, 1999). Critically, conceptual priming was identified for meaningful words but not
meaningless words in two behavioral tasks.
Two behavioral tasks were used to index conceptual priming for uncommon words that
were used as stimuli in the recognition paradigm. Our a priori hypothesis was that conceptual
priming could occur for meaningful words but not meaningless words, and key contrasts thus
focused on differences in conceptual priming for repeat stimuli compared to novel stimuli within
these meaningfulness categories. Behavioral data were collected from 12 individuals (5 male,
ages 18 to 24).
In a lexical decision task, subjects rapidly categorized repeat and novel meaningful and
meaningless uncommon words into “word” and “not word” categories. When uncommon words are included, this categorization judgment requires access to pertinent conceptual information
(i.e., uncommon word meaning), and would be speeded to the extent that access to associated
conceptual meaning was enhanced by conceptual priming. During encoding, 60 uncommon
words (selected at random for each subject; 2000 ms presentation, 2000-3500 ms randomized
ISI) were divided into meaningful and meaningless categories by ratings as in the recognition paradigm. During the lexical decision test, all 60 uncommon words were presented again along with 60 novel uncommon words (selected at random for each subject) and 60 common, high- frequency words (e.g. table, lamp, church), in randomized order (1000 ms presentation, 2000-
3000 ms randomized ISI). Subjects pressed one button to indicate that the item on the screen was 107 a word, and another button to indicate that the item was not a word. Speed and accuracy were
emphasized.
Fluency of processing is thought to lead to an enhancement of positive affect
(Winkelman, Halberstadt, Fazendeiro, & Catty, 2006). Thus, conceptual priming was expected to
lead to an increase in liking ratings in a task that used a scale to indicate the extent to which each stimulus was liked at a gut level. During encoding, 60 uncommon words (selected at random for each subject; 2000 ms presentation, 2000-3500 ms randomized ISI) were divided into
meaningful and meaningless categories using ratings as in the recognition paradigm. During the
liking rating test, all 60 words were presented again along with 60 novel uncommon words
(selected at random for each subject), in randomized order (1000 ms presentation, 2000-3000 ms
randomized ISI). Subjects rated each stimulus using a 4-point scale including the response
categories: “dislike the most,” “like the least,” “like a little more,” and “like the most.” Adjacent
ratings were assigned to adjacent fingers, and the button assigned to the lowest rating on the
liking scale was the button assigned to the highest rating on the meaningfulness scale, such that
high meaningfulness ratings would not lead to higher liking ratings by virtue of response
priming. Subjects were cued to make ratings at a gut level, and response speed was not
emphasized.
Following the lexical decision and liking rating tasks (the order of which was
counterbalanced across subjects), all uncommon word stimuli from each task were again
categorized into meaningful and meaningless categories. This allowed us to categorize novel
stimuli during the test phase of each task into meaningful and meaningless categories. For repeat
stimuli, ratings made following the two tasks were highly consistent with the initial rating made 108 during encoding for each task (94% of stimuli were assigned to the same meaningfulness category on average, SE=4.3%), thus indicating that classification of novel words was valid.
Table 5 summarizes data from the lexical decision and liking rating tasks. Overall, 5.9%
(SE=2.2%) of stimuli were endorsed as real words and were excluded from analyses, 50.2%
(SE=2.3%) were categorized as meaningful, and 43.9% (SE=2.4%) as meaningless. In the lexical decision task, accuracy in endorsing uncommon words as nonwords was high, and did not differ for meaningful and meaningless old and new stimuli [F(1,11)=0.003, p=0.96). However, reaction times did differ for these categories [F(1,11)=4.4, p=0.06]. Conceptual priming was identified selectively for meaningful stimuli in that responses to old meaningful words were faster than to new meaningful words [t(11)=2.4, p=0.03]. In contrast, reaction times to meaningless old and new words did not differ significantly [t(11)=1.0, p=0.32]. In the liking rating task, old and new words were rated differently for each meaningfulness category [F(1,11)=4.9, p=0.05].
Meaningful old words were given higher ratings than were meaningful new words [t(11)=2.3, p=0.04], and old meaningless words were not given higher ratings than new meaningful words
[t(11)=0.4, p=0.69]. Critically, the old/new comparison yielded equivalent perceptual priming for meaningful and meaningless categories, which were subject to similar perceptual processing.
Furthermore, the physical features of stimuli in the meaningful and meaningless categories were equivalent (all uncommon words), and thus old/new response differences must have instead been based upon differential access to associated conceptual meaning. Thus, in both tasks, conceptual priming was identified for meaningful but not meaningless stimuli. These findings indicate that the meaningful and meaningless categories varied in perceived meaningfulness and varied in their ability to support conceptual priming. 109 With regard to explicit memory, high-confidence know responses to words in the meaningful and meaningless categories indexed veridical recognition memory, as indicated by significantly positive d′ values (Table 4). Critically, two lines of evidence indicate that these categories were matched in familiarity strength. Whereas the overall pattern of recognition responses differed between meaningfulness categories, quantities of high-confidence know responses did not differ, indicating that these responses were insensitive to inferred meaning.
Because recollection is frequently thought to entail simultaneous familiarity (Yonelinas, 2002), familiarity was undoubtedly greater for the entire sets of meaningful compared to meaningless words in that meaningful words were recognized with a greater proportion of remember responses. Nevertheless, by focusing on know responses that occured in the absence of recollection, it was possible to isolate recognition responses that reflected a single interval of familiarity strength unaffected by the confounding influence of recollection on memory strength.
Thus, meaningful and meaningless words differed in the degree to which they supported conceptual priming, and subsets of these categories (words recognized with high-confidence know responses) were matched in explicit memory strength.
Neural correlates of conceptual priming were obtained by contrasting ERP responses to meaningful vs. meaningless words that were all recognized with high-confidence familiarity responses (Figure 20). FN400 potentials appeared to be more positive for meaningful compared to meaningless words at mid-frontal recording electrodes. Formal comparisons made over seven scalp regions (Figure 19) and three latency intervals that captured the FN400 (300-500 ms) as well as later potentials (500-700 and 700-900 ms) substantiated this observation. The meaningful/meaningless contrast differed significantly over regions and latency intervals [3-way interaction F(3.1,53.3)=3.12, p=0.03], due to significant positive enhancements at mid-frontal 110 electrodes over all intervals [condition main effect F(1,17)=8.86, p=0.009] but not at other
regions over any interval (p’s>0.15 for all main effects and interactions involving condition).
Meaningfulness and the associated phenomena of conceptual priming thus were associated with
mid-frontal ERP enhancements that encompassed FN400 and subsequent potentials.
ERP differences between meaningful and meaningless words attributed here to
conceptual priming could have merely resulted from an influence of inferred meaning on ERP
amplitude instead of repetition-enhanced access to word meaning. If so, such differences would
be evident regardless of whether words were viewed for the first or second time. During
encoding, ERP correlates of meaningful and meaningless words did not differ in amplitude for
any of the seven regions and three latency intervals examined in the recognition analysis (all
p’s>0.20). Thus, ERP differences between meaningful and meaningless words identified during
recognition testing indeed reflected distinct retrieval processes engaged for these two categories.
Direct evidence in favor of the hypothesis that FN400 potentials obtained during
recognition testing reflect conceptual priming instead of familiarity would include reliable
positive enhancements of these potentials for recognition of meaningful, but not meaningless,
words. To identify ERP correlates of recognition memory, ERPs to repeated meaningful and
meaningless words all correctly endorsed with high-confidence know responses were compared
to ERPs to missed repeated words that were incorrectly endorsed either as new items or endorsed with medium- or low-confidence know responses (both for which discrimination sensitivity was no greater than chance; Table 4). ERP results shown in Figure 21 included FN400 recognition memory effects (300-500 ms) that were reliable at mid-frontal and right-frontal regions for
meaningful words (Table 6). In contrast, no recognition memory effects at frontal regions were
present for meaningless words (Table 6). ERP correlates of recognition memory at posterior 111 scalp sites were reliable for both meaningful and meaningless words at later latency intervals.
For meaningful words, significant positive enhancements at central, and right-, middle-, and left- posterior regions extended from 300-900 ms whereas positive enhancements at these regions were significant for meaningless words only from 500-700 ms (Table 6).
34. Discussion
Difficulties associated with separating neural correlates of familiarity from those of conceptual priming were surmounted by using stimuli for which these forms of memory were somewhat orthogonal. Meaningful words were capable of supporting conceptual priming whereas meaningless words were not. In addition, explicit memory was stronger for meaningful than meaningless words. However, subsets of meaningful and meaningless items were endorsed with high-confidence recognition responses, and were therefore equated in explicit memory strength. By focusing on these items, neural correlates of conceptual priming could be obtained
in isolation from neural correlates of familiarity. Conceptual priming ERP correlates included
FN400 potentials and subsequent potentials with an identical scalp distribution. Furthermore,
neural correlates of familiarity-based recognition included FN400 potentials only for meaningful
words, for which conceptual priming was also indexed by such potentials. In contrast, potentials
at posterior scalp electrodes from 500 to 700 ms were associated with familiarity-based
recognition for both meaningful and meaningless words, indicating that ERP correlates of
familiarity resembled those commonly attributed to recollection (Friedman & Johnson, 2000;
Paller, 2000). This pattern of results detracts from the evidence supporting dual-process models
of explicit memory by indicating that dissociations between electrophysiological correlates of
familiarity and recollection (reviewed in M.D. Rugg & Curran, in press) may have arose 112 artificially due to the misidentification of neural correlates of conceptual priming as those of
familiarity.
High-confidence familiarity responses were quantitatively similar for both meaningful
and meaningless stimuli, and this was taken as evidence that explicit memory strength was
matched for these categories. We thus interpreted corresponding ERP differences as neural correlates of conceptual priming. An alternative possibility, however, is that familiarity varied qualitatively between meaningful and meaningless items, and this qualitative difference was responsible for the observed ERP differences. We are unaware, however, of any theoretical accounts indicating that familiarity varies qualitatively as a function of stimulus meaningfulness.
Furthermore, we consider such a distinction to be unlikely given that a defining characteristic of familiarity is an absence of contextual retrieval that would seemingly be required to support qualitative conceptual-based differences.
Several previous studies also have employed minimally-meaningful stimuli, including pseudowords (Curran, 1999) and abstract geometrical shapes (Curran et al., 2002; Groh-Bordin et al., 2006), to obtain neural correlates of familiarity that were assumed to be uncontaminated by neural correlates of conceptual priming (see Section 17). ERP results were used to associate
FN400 potentials with familiarity, and were thus taken as support for the dual-process account. It is very likely, however, that subjects attempted to find meaning in these purportedly meaningless stimuli, especially given that doing so could have aided subsequent recognition. Indeed, the experiments described in Part Six indicated that abstract geometric shapes could be perceived as meaningful, and that conceptual priming was indexed by FN400 potentials under these circumstances (Voss & Paller, 2007). The present stimuli were also minimally meaningful, and results indicate that FN400 correlates of conceptual priming can contaminate 113 electrophysiological correlates of recognition, and that this can occur for verbal materials akin to
those commonly employed in memory research.
Electrophysiological correlates of recognition memory were obtained here using stimuli
that had been previously viewed (i.e., old hits vs. misses). FN400 correlates of conceptual
priming were nonetheless identified for recognition of meaningful words, and this argues in
favor of the view that conceptual priming is not an entirely automatic process engaged for all
recently-encountered stimuli. Instead, conceptual priming can be influenced by many of the same variables that influence recognition (reviewed in Yonelinas, 2002), and is therefore likely to be present during recognition tests especially for the stimuli that are recognized with the greatest confidence or accuracy. Furthermore, during latencies subsequent to FN400 potentials, meaningful words produced enhanced amplitudes compared to meaningless words at mid-frontal electrodes, but not compared to the baseline condition of missed words. Thus, conceptual priming was associated with FN400 potentials as well as subsequent potentials with an identical scalp distribution (Figure 20), yet contaminated neural correlates of familiarity-based recognition primarily during the FN400 latency (Figure 21). Similar results were obtained in Part Six, when novel items, instead of repeated missed items, were used as a recognition memory baseline (Voss
& Paller, 2007).
Neuroanatomical support for dual-process models has been obtained from other sources, although this evidence is contentious. Many findings in amnesic patients accommodate the dual- process perspective (e.g. Duzel, Vargha-Khadem, Heinze, & Mishkin, 2001; Holdstock et al.,
2002; Yonelinas et al., 2002) by suggesting that recollection depends on the integrity of the hippocampus whereas familiarity is supported by the adjacent neocortex of the parahippocampal gyrus (Aggleton & Brown, 2006). Other evidence casts serious doubt on this simple dichotomy, 114 however, by indicating that the hippocampus is critical for both recollection and familiarity
(Wais et al., 2006; Wixted & Squire, 2004). In healthy subjects, fMRI measures have often indicated that recollection recruits greater activity within the hippocampus than does familiarity, whereas familiarity is more directly tied to activity within adjacent neocortex (e.g. Davachi et al.,
2003; Eldridge et al., 2000; Montaldi, Spencer, Roberts, & Mayes, 2006; Ranganath et al., 2004;
Yonelinas et al., 2005). However, patterns of activity associated with both recollection and familiarity often encompass both the hippocampus and adjacent neocortex to different degrees for each memory type, again casting doubt on any clear dichotomy between corresponding neural substrates. Furthermore, a possibility suggested by the present results and by a critical examination of the ERP literature (Paller et al., in preparation) is that fMRI correlates of recognition reflect other co-occurring memory processes. Some progress is being made in validating fMRI correlates of recognition (e.g. Daselaar, Fleck, & Cabeza, 2006), although future studies should consider a wide variety of mnemonic phenomena, including conceptual priming.
115 Part Eight: Concluding Remarks
It is now possible to revisit the two central controversies presented in Section 15 with new perspectives provided by results from the experiments described in Parts Four through
Seven.
Controversy 1: The functional relationship between familiarity and conceptual priming
The majority of the evidence provided here suggests that familiarity and conceptual priming are functionally independent. If familiarity resulted from conceptual priming, one would expect that neural markers of the two should always co-occur. For famous faces, ERP and fMRI correlates of each memory phenomena could be dissociated during the performance of a single conceptual priming task. Similarly, ERP correlates of each could be dissociated using minimalist visual shapes and uncommon words during recognition testing. These results converge with the finding that these memory processes can be dissociated in amnesia (Levy et al., 2004), and pose serious difficulties for any theory proposing a tight coupling between familiarity and conceptual priming. Furthermore, that these results were obtained using a wide range of stimuli (faces, squiggles, words) and testing conditions (implicit, explicit) indicates that results are likely to generalize to other experimental situations.
If FN400 potentials index conceptual priming and later-onset posterior-centered (LPC) potentials index familiarity, one might ask “If conceptual priming and familiarity are functionally independent, why then are FN400 potentials and LPC potentials often identified together during recognition?” The answer is that conceptual priming and familiarity, while not functionally related, tend to co-occur during recognition testing whenever stimuli can support both memory phenomena. For instance, the repetition of a meaningful stimulus, such as a word 116 or nameable picture, will engender both processes. This is especially the case because familiarity and conceptual priming are influenced in a parallel fashion by a variety of experimental parameters (Part 3). This implies that it is highly likely that the best-remembered items during a recognition test will also be subject to the greatest level of conceptual priming, thus biasing neural correlates of recognition to include measures of both explicit memory (LPC) and conceptual priming (FN400).
The strategy employed here to tease apart neural correlates of each process was to use novel behavioral paradigms to create conditions in which familiarity and conceptual priming were somewhat orthogonal, and to include corresponding behavioral measures in order to substantiate distinctions. In doing so, it was possible to dissociate them and identify circumstances in which each process, and their corresponding neural correlates, occurred in isolation (e.g. Figures 8cd, 11, 17, 20, and 21). Further evidence for their functional independence also comes indirectly from previous ERP studies, which have often identified obvious across-subject relationships between recognition performance and LPC potentials (as in
Figure 18), but have seldom found even hints of weak relationships between recognition and
FN400 (e.g., Curran 2006). Weak relationships that have been occasionally identified between recognition performance and FN400 can be dismissed when considering both the strong relationship identified here between conceptual priming and FN400 (Section 21) and the previous argument that they tend to co-occur under some circumstances.
Controversy 2: Dual-process vs. single-process models of recognition The finding that
FN400 potentials can reflect conceptual priming poses serious difficulties for dual-process models of recognition. These models propose that separate recollection and familiarity processes can provide unique contributions to recognition, and give rise to the phenomenological 117 experiences of recollection and familiarity, respectively. In contrast, single-process theories
propose that a single global memory strength variable drives recognition and encompasses both
recollection and familiarity.
Currently, there is no consensus over whether behavioral data in healthy individuals
accommodate one or two retrieval processes (Wixted, 2007; Wixted & Stretch, 2004; Yonelinas,
2002). This debate can be condensed into an argument over whether signal detection models which propose two variables provide a better fit to behavioral data than do models which propose only one variable. Ultimately, however, this debate is irrelevant in that its resolution cannot answer the question of whether familiarity and recollection represent unique brain processes. It is not sufficient to determine which model provides the best fit to behavioral data, in that there is no guarantee that nature has devised the most logical, parsimonious, or otherwise
“best” solution to the problem of recognition. Instead, it is necessary to examine the neural substrates of recollection and familiarity and determine whether they are identical or unique.
As reviewed in Section 34, neither investigations of amnesic patients nor those using fMRI in healthy individuals have provided clear dissociations between recollection and familiarity in dependence on the hippocampus and surrounding cortex, respectively. The centerpiece of the neural evidence in support of dual-process models has been the so-called
“double-dissociation” between ERP correlates of recollection and familiarity, whereby recollection is indexed by LPC potentials and familiarity by FN400 potentials. However, the results presented here indicate that FN400 has been erroneously assigned to familiarity, and is instead closely related to conceptual priming. Previous reports have thus provided double- dissociations between neural correlates of conceptual priming and those of explicit memory
(both recollection and familiarity). I argue that familiarity is indexed by LPC potentials with 118 reduced amplitudes compared to recollection. Thus, recollection and familiarity vary quantitatively, not qualitatively, and electrophysiological evidence supports single-process models.
As a final note, it is curious that dual-process models have been so widely accepted despite a profound lack of convincing evidence. Indeed, even without the empirical findings presented here, the literature review provided in Part 3 makes it clear that the strongest existing evidence in favor of dual-process models, the ephemeral relationship between FN400 and familiarity, was shaky at best. It is my view that this premature acceptance of dual-process models provides a striking example of how enthusiasm for appealing theories can cloud appropriate scientific rigor. In this case, it is tempting to believe that distinct phenomenological experiences should be supported by distinct retrieval processes. This has caused many investigators to mistakenly overlook the paucity of the evidence, and even to bend the evidence in favor of dual-process models (e.g., Curran, 2004; Woodruff et al., 2006).
Toward a unified model of long-term memory I will preface the final portion of this discussion by stating that the following remarks are highly speculative, and are presented here only to stimulate consideration that could lead to empirical testing. I have presented evidence for dissociations between familiarity and conceptual priming. However, this does not indicate that these memory phenomena act upon distinct memory traces. For instance, if a chair is encoded into memory and later memory events include familiarity-based recognition and conceptual priming, it is not necessary that two separate representations of the chair exist somewhere in the brain, one that is primed and one that is recognized. Instead, these memory phenomena likely represent dissociable retrieval processes that act on the same memory trace. How might a memory system be organized in order to support such function? 119 The hippocampus-dependent long-term declarative memory system is distinctive from other memory systems in its purpose—it allows us to mentally time-travel into the past, arguably so that we may better prepare for the future. But how is this possible? I propose that three processes are involved, which I will term “recapitulation,” “indexing,” and “decision.” These processes encompass the memory phenomena of priming and declarative memory in the following ways.
Compelling evidence indicates that mental-imagery and vivid-recollection of particular stimuli produce patterns of neural activity similar to those produced when the same stimuli are perceived (Kosslyn, Thompson, Kim, & Alpert, 1995; Wheeler et al., 2006). This implies that a potential mechanism of mental-time travel is neural time-travel. Memory for a stimulus could involve the recapitulation of the initial events of the stimulus’ perception. I propose that the recapitulation process is the series of neural processing steps that leads to the recapitulation of encoding conditions. Furthermore, all qualities of long-term memory expression, including perceptual priming, conceptual priming, familiarity, and recollection, involve the recapitulation process. What sets these distinct expressions of long-term memory apart is the extent to which the recapitulation process occurs and the extent to which the indexing process is also involved, by the following scenario.
When the sensory information induced by perception of a novel stimulus is processed by the brain, it can produce a lasting memory trace in all activated neural systems. Of course, the systems involved will depend largely on the nature of the stimulus. This selective involvement extends beyond obvious distinctions to include complex relationships between systems. For instance, an entirely meaningless visual stimulus, such as a novel kaleidoscope image, would likely produce activity compartmentalized in visual cortex. However, the auditory system would 120 be involved to the extent to which a purely visual stimulus induced auditory imagery (e.g., seeing
a picture of a cow and thinking “moo”). Similarly, complex meaning-based representations that
likely exist in the furthest reaches of the ventral visual processing stream (anterior temporal lobes) would be created to the extent to which the stimulus was meaningful or cued associations with other meaningful stimuli.
The indexing process attempts to actively support time-travel by creating an index with which these traces can be later revived. While the sheer number of possible events makes for a combinatorial nightmare, the declarative memory system appears to be tailored to solving exactly such a task. For instance, the hippocampus and surrounding neocortex receive terminal inputs from every specialized sensory system, and perform exactly the types of computations that
would be necessary to create unique indices for complex patterns of neural activity [e.g., decorrelation of inputs (Leutgeb, Leutgeb, Moser, & Moser, 2007)]. The hippocampus and surrounding neocortex are thus the neural machinery that primarily supports indexing, especially when the pattern of input neural activity extends across multiple sensory systems that lack direct neural connections. It is likely that cortical association areas also support indexing in unique ways. For instance, the frontal lobe may be involved in indexing especially when the incoming
stimulus involves neural representations of language or other complex spatial or temporal
sequences.
In this hypothesized system, the conscious awareness of remembering does not result via
the activity of a specialized process per se, but rather occur as a byproduct of the recapitulation
process. Mounting evidence suggests that conscious awareness is a phenomenon directly tied the
qualitative nature of neural processing, such as coherent processing across modality-specific
neural regions (Srinivasan, Russell, Edelman, & Tononi, 1999) or recurrent sensory processing 121 (Lamme & Roelfsema, 2000). I thus consider the awareness of remembering to entail high- fidelity recapitulation of processing within a single sensory modality (weak or dull awareness) or across multiple sensory modalities (strong or vivid awareness).
I will now consider how these pieces interact to support memory. When a stimulus is viewed for the second time, a pattern of neural activity is produced which overlaps with the pattern that was produced upon initial viewing. The degree of overlap is the degree of recapitulation, and could be influenced by a variety of endogenous and exogenous factors, such as the similarity between viewing conditions (e.g. lighting or viewpoint), cognitive states, retrieval effort, etc.
When recapitulation is weak, this could result in less or facilitated neural processing, in that information is processed within a neural circuit that has been potentiated by the initial processing. Critically, weak recapitulation involves recapitulation of initial activity within one unimodal cortical region or weak concerted replay across several unimodal cortical regions. This sort of activity constitutes the neural processing that gives rise to behavioral priming phenomena.
Priming is thus recapitulated processing that does not trigger awareness simply because the nature of the neural processing (i.e., not concerted activity across modality-specific regions or strong activity within a given region) is not of the kind required to engender awareness (although it is possible that a very minimal awareness could occur). This theory does not make a strong distinction between the behavioral phenomena of perceptual and conceptual priming, but suggests that they differ merely in the extent to which the stimuli for which they occur elicit activity in a segregated perceptual system (such as a meaningless abstract visual shape) vs. across several systems or including meaning-related representations (such as an abstract visual shape that can be readily given a verbal label). Weak recapitulation can also involve facilitated 122 neural processing of activity related to the indexing process. Thus, neural correlates of priming could include facilitated responses within the cortical regions involved in initial processing (i.e., visual cortex) or indexing (i.e., medial temporal lobe cortex or frontal cortex).
When recapitulation is stronger, it can involve the reactivation of an index, such that a greater fraction of the initial processing conditions can be reinstated. When the initial conditions included concerted activity across cortical regions, this is more likely to result in the awareness of remembering, or conscious recollection. When the initial conditions did not include concerted activity (either due to the nature of the stimulus or other factors that could reduce the extent of processing, such as attention), this is more likely to result in a weak awareness of remembering, or the experience of familiarity without recollection. In either case, processing is fundamentally distinct from that required to produce priming without awareness. However, it is possible that weak recapitulation could precede stronger recapitulation or occur simultaneously in neural regions segregated from those involved in stronger recapitulation. My use of the term
“awareness” connotes a “reliving” experience that is fundamentally different from the
“awareness that an item is old,” which involves a decision making process.
The recapitulation and indexing processes operate independently from the decision process, which is what is generally measured in tests of explicit memory. The decision process represents the processing required to make the determination that something has been experienced in the past. It is difficult to avoid homuncular reasoning when accounting for decision, but I conceive of decision as a relatively automatic process that derives directly from the recapitulation and indexing neural processing that leads to awareness. Under this scenario, the stronger the awareness, either absolutely or relative to some baseline such as a novel stimulus, the more likely the decision process will point to “old.” This can occur irrespective of 123 whether the awareness stems from a weak recapitulation process that produces weak awareness
[such as when priming biases recognition decisions (Rajaram & Geraci, 2000; Verfaellie &
Cermak, 1999; Voss, Baym, & Paller, in preparation; Wolk et al., 2005)] or from strong awareness resulting from strong recapitulation coupled with indexing processing.
It would be difficult to test key predictions of this preliminary model in humans given the
resolution of current neuroimaging tools. For instance, ERP and fMRI methods permit the measurement of only large populations of neurons. Thus, it is likely that indexing operations are commonly identified in neuroimaging experiments, whereas recapitulation within small-scale, item-specific networks would remain largely invisible.
Of course, this tentative model is but one of many that could account for the current empirical findings regarding long-term memory. Ultimately, a unified model of the component processes of memory must account for memory-related neural processing, and specify how this neural processing results in memory behaviors and corresponding conscious experiences.
124 Bibliography
Aggleton, J. P., & Brown, M. W. (2006). Interleaving brain systems for episodic and recognition memory. Trends in Cognitive Sciences, 10(10), 455-463.
Allison, T., Puce, A., Spencer, D. D., & McCarthy, G. (1999). Electrophysiological studies of human face perception. I: Potentials generated in occipitotemporal cortex by face and non-face stimuli. Cerebral Cortex, 9(5), 415-430.
Azimian-Faridani, N., & Wilding, E. L. (2006). The influence of criterion shifts on electrophysiological correlates of recognition memory. Journal of Cognitive Neuroscience, 18(7), 1075-1086.
Bishop, K. I., & Curran, T. (1998). An investigation of the effects of benzodiazepine receptor ligands and of scopolamine on conceptual priming. Psychopharmacology, 140, 345-353.
Brewer, J. B., Zhao, Z., Desmond, J. E., Glover, G. H., & Gabrieli, J. D. (1998). Making memories: brain activity that predicts how well visual experience will be remembered. Science, 281(5380), 1185-1187.
Brown, M. W., & Aggleton, J. P. (2001). Recognition memory: what are the roles of the perirhinal cortex and hippocampus? Nature Reviews Neuroscience, 2(1), 51-61.
Buckner, R. L., Koutstaal, W., Schacter, D. L., & Rosen, B. R. (2000). Functional MRI evidence for a role of frontal and inferior temporal cortex in amodal components of priming. Brain, 123(3), 620-640.
Burock, M. A., Buckner, R. L., Woldorff, M. G., Rosen, B. R., & Dale, A. M. (1998). Randomized event-related experimental designs allow for extremely rapid presentation rates using functional MRI. Neuroreport, 9(16), 3735-3739.
Cameron, K. A., Yashar, S., Wilson, C. L., & Fried, I. (2001). Human hippocampal neurons predict how well word pairs will be remembered. Neuron, 30(1), 289-298.
Challis, B. H., & Sidhu, R. (1993). Dissociative effect of massed repetition on implicit and explicit measures of memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19(1), 115-127.
Cox, R. W. (1996). AFNI: software for analysis and visualization of functional magnetic resonance neuroimages. Computers and Biomedical Research, 29(3), 162-173. 125 Curran, T. (1999). The electrophysiology of incidental and intentional retrieval: ERP old/new effects in lexical decision and recognition memory. Neuropsychologia, 37(7), 771-785.
Curran, T. (2000). Brain potentials of recollection and familiarity. Memory & Cognition, 28(6), 923-938.
Curran, T. (2004). Effects of attention and confidence on the hypothesized ERP correlates of recollection and familiarity. Neuropsychologia, 42(8), 1088-1106.
Curran, T., & Cleary, A. M. (2003). Using ERPs to dissociate recollection from familiarity in picture recognition. Brain Research Cognitive Brain Research, 15(2), 191-205.
Curran, T., DeBuse, C., Woroch, B., & Hirshman, E. (2006). Combined pharmacological and electrophysiological dissociation of familiarity and recollection. The Journal of Neuroscience, 26(7), 1979-1985.
Curran, T., & Dien, J. (2003). Differentiating amodal familiarity from modality-specific memory processes: an ERP study. Psychophysiology, 40(6), 979-988.
Curran, T., & Hancock, J. (in press). The FN400 indexes familiarity-based recognition of faces. Neuroimage.
Curran, T., Schacter, D. L., Johnson, M. K., & Spinks, R. (2001). Brain potentials reflect behavioral differences in true and false recognition. Journal of Cognitive Neuroscience, 13(2), 201-216.
Curran, T., Tanaka, J. W., & Weiskopf, D. M. (2002). An electrophysiological comparison of visual categorization and recognition memory. Cognitive, Affective, and Behavioral Neuroscience, 2(1), 1-18.
Curran, T., Tepe, K. L., & Piatt, C. (2006). ERP explorations of dual processes in recognition memory. In H. D. Zimmer, A. Mecklinger & U. Lindenberger (Eds.), Handbook of binding and memory: perspectives from cognitive neuroscience (pp. 467-492). Oxford: Oxford University Press.
Cycowicz, Y. M., Friedman, D., & Snodgrass, J. G. (2001). Remembering the color of objects: an ERP investigation of source memory. Cerebral Cortex, 11(4), 322-334.
D'Esposito, M. (in press). Short term and working memory systems. In J. Byrne (Ed.), Learning and memory: a comprehensive reference. Oxford: Elsevier. 126 Daselaar, S. M., Fleck, M. S., & Cabeza, R. (2006). Triple dissociation in the medial temporal lobes: recollection, familiarity, and novelty. Journal of Neurophysiology, 96(4), 1902-1911.
Davachi, L., Mitchell, J. P., & Wagner, A. D. (2003). Multiple routes to memory: distinct medial temporal lobe processes build item and source memories. Proceedings of the National Academy of Sciences of the United States of America, 100(4), 2157-2162.
Diana, R. A., Vilberg, K. L., & Reder, L. M. (2005). Identifying the ERP correlate of a recognition memory search attempt. Brain Research Cognitive Brain Research, 24(3), 674-684.
Dobbins, I. G., Schnyer, D. M., Verfaellie, M., & Schacter, D. L. (2004). Cortical activity reductions during repetition priming can result from rapid response learning. Nature, 428(6980), 316-319.
Donaldson, D. I., Petersen, S. E., & Buckner, R. L. (2001). Dissociating memory retrieval processes using fMRI: evidence that priming does not support recognition memory. Neuron, 31(6), 1047-1059.
Duarte, A., Ranganath, C., Winward, L., Hayward, D., & Knight, R. T. (2004). Dissociable neural correlates for familiarity and recollection during the encoding and retrieval of pictures. Brain Research Cognitive Brain Research, 18(3), 255-272.
Duzel, E., Habib, R., Schott, B., Schoenfeld, A., Lobaugh, N., McIntosh, A. R., et al. (2003). A multivariate, spatiotemporal analysis of electromagnetic time-frequency data of recognition memory. Neuroimage, 18(2), 185-197.
Duzel, E., Vargha-Khadem, F., Heinze, H. J., & Mishkin, M. (2001). Brain activity evidence for recognition without recollection after early hippocampal damage. Proceedings of the National Academy of Sciences of the United States of America, 98(14), 8101-8106.
Duzel, E., Yonelinas, A. P., Mangun, G. R., Heinze, H. J., & Tulving, E. (1997). Event-related brain potential correlates of two states of conscious awareness in memory. Proceedings of the National Academy of Sciences of the United States of America, 94(11), 5973-5978.
Eldridge, L. L., Knowlton, B. J., Furmanski, C. S., Bookheimer, S. Y., & Engel, S. A. (2000). Remembering episodes: a selective role for the hippocampus during retrieval. Nature Neuroscience, 3(11), 1149-1152.
Eldridge, L. L., Sarfatti, S., & Knowlton, B. J. (2002). The effect of testing procedure on remember-know judgments. Psychonomic Bulletin & Review, 9(1), 139-145. 127 Engel, A. K., Moll, C. K., Fried, I., & Ojemann, G. A. (2005). Invasive recordings from the human brain: clinical insights and beyond. Nature Reviews Neuroscience, 6(1), 35-47.
Fabiani, M., Karis, D., & Donchin, E. (1986). P300 and recall in an incidental memory paradigm. Psychophysiology, 23(3), 298-308.
Fabiani, M., Karis, D., & Donchin, E. (1990). Effects of mnemonic strategy manipulation in a Von Restorff paradigm. Electroencephalography and Clinical Neurophysiology, 75(2), 22-35.
Fell, J., Klaver, P., Lehnertz, K., Grunwald, T., Schaller, C., Elger, C. E., et al. (2001). Human memory formation is accompanied by rhinal-hippocampal coupling and decoupling. Nature Neuroscience, 4(12), 1259-1264.
Fernandez, G., Klaver, P., Fell, J., Grunwald, T., & Elger, C. E. (2002). Human declarative memory formation: segregating rhinal and hippocampal contributions. Hippocampus, 12(4), 514- 519.
Forman, S. D., Cohen, J. D., Fitzgerald, M., Eddy, W. F., Mintun, M. A., & Noll, D. C. (1995). Improved assessment of significant activation in functional magnetic resonance imaging (fMRI): use of a cluster-size threshold. Magnetic Resonance in Medicine, 33(5), 636-647.
Fortin, N. J., Wright, S. P., & Eichenbaum, H. (2004). Recollection-like memory retrieval in rats is dependent on the hippocampus. Nature, 431(7005), 188-191.
Friedman, D. (1990). ERPs during continuous recognition memory for words. Biological Psychology, 30(1), 61-87.
Friedman, D. (2004). ERP studies of recogntion memory: differential effects of familiarity, recollection, and episodic priming. Cognitive Sciences, 1, 81-121.
Friedman, D., & Johnson, R., Jr. (2000). Event-related potential (ERP) studies of memory encoding and retrieval: a selective review. Microscopy Research and Technique, 51(1), 6-28.
Gabrieli, J. D. (1998). Cognitive neuroscience of human memory. Annual Review of Psychology, 49, 87-115.
Gardiner, J. M., & Java, R. I. (1991). Forgetting in recognition memory with and without recollective experience. Memory & Cognition, 19(6), 617-623.
Gershberg, F. B. (1997). Implicit and explicit conceptual memory following frontal lobe damage. Journal of Cognitive Neuroscience, 9(1), 105-116. 128 Golding, J. M., & MacLeod, C. M. (1998). Intentional forgetting: interdisciplinary approaches. Mahwah, N.J.: L. Erlbaum Associates.
Gonsalves, B., & Paller, K. A. (2000a). Brain potentials associated with recollective processing of spoken words. Memory & Cognition, 28(3), 321-330.
Gonsalves, B., & Paller, K. A. (2000b). Neural events that underlie remembering something that never happened. Nature Neuroscience, 3(12), 1316-1321.
Graf, P., Shimamura, A. P., & Squire, L. R. (1985). Priming across modalities and priming across category levels: extending the domain of preserved function in amnesia. Journal of Experimental Psychology: Learning, Memory, and Cognition, 11(2), 386-396.
Graf, P., Squire, L. R., & Mandler, G. (1984). The information that amnesic patients do not forget. Journal of Experimental Psychology: Learning, Memory, and Cognition, 10(1), 164-178.
Grill-Spector, K. (in press). Priming. In J. Byrne (Ed.), Learning and memory: a comprehensive reference. Oxford: Elsevier.
Groh-Bordin, C., Zimmer, H. D., & Ecker, U. K. (2006). Has the butcher on the bus dyed his hair? When color changes modulate ERP correlates of familiarity and recollection. Neuroimage.
Groh-Bordin, C., Zimmer, H. D., & Mecklinger, A. (2005). Feature binding in perceptual priming and in episodic object recognition: evidence from event-related brain potentials. Brain research, 24(3), 556-567.
Grunwald, T., Lehnertz, K., Heinze, H. J., Helmstaedter, C., & Elger, C. E. (1998). Verbal novelty detection within the human hippocampus proper. Proceedings of the National Academy of Sciences of the United States of America, 95(6), 3193-3197.
Guillem, F., Bicu, M., & Debruille, J. B. (2001). Dissociating memory processes involved in direct and indirect tests with ERPs to unfamiliar faces. Brain Research Cognitive Brain Research, 11(1), 113-125.
Guillem, F., Rougier, A., & Claverie, B. (1999). Short- and long-delay intracranial ERP repetition effects dissociate memory systems in the human brain. Journal of Cognitive Neuroscience, 11(4), 437-458.
Guo, C., Duan, L., Li, W., & Paller, K. A. (2006). Distinguishing source memory and item memory: brain potentials at encoding and retrieval. Brain Research, 1118(1), 142-154. 129 Halgren, E., & Smith, M. E. (1987). Cognitive evoked potentials as modulatory processes in human memory formation and retrieval. Human Neurobiology, 6(2), 129-139.
Hamann, S. B. (1990). Level-of-processing effects in conceptually driven implicit tasks. Journal of Experimental Psychology: Learning, Memory, and Cognition, 16(6), 970-977.
Hamann, S. B., & Squire, L. R. (1997). Intact perceptual memory in the absence of conscious memory. Behavioral Neuroscience, 111(4), 850-854.
Heit, G., Smith, M. E., & Halgren, E. (1988). Neural encoding of individual words and faces by the human hippocampus and amygdala. Nature, 333(6175), 773-775.
Henson, R. N. (2003). Neuroimaging studies of priming. Progress in Neurobiology, 70(1), 53- 81.
Henson, R. N., Cansino, S., Herron, J. E., Robb, W. G., & Rugg, M. D. (2003). A familiarity signal in human anterior medial temporal cortex? Hippocampus, 13(2), 301-304.
Hintzman, D. L., & Curran, T. (1994). Retrieval dynamics of recognition and frequency judgments. Journal of Memory and Language, 33, 1-18.
Hirshman, E., Fisher, J., Henthorn, T., Arndt, J., & Passannante, A. (2002). Midazolam amnesia and dual-process models of the word-frequency mirror effect. Journal of Memory and Language, 47, 499-516.
Holdstock, J. S., Mayes, A. R., Roberts, N., Cezayirli, E., Isaac, C. L., O'Reilly, R. C., et al. (2002). Under what conditions is recognition spared relative to recall after selective hippocampal damage in humans? Hippocampus, 12(3), 341-351.
Jacoby, L. L., & Dallas, M. (1981). On the relationship between autobiographical memory and perceptual learning. Journal of Experimental Psychology: General, 110(3), 306-340.
James, W. (1890). The principles of psychology. New York,: H. Holt and company.
Johnson, M. K., Kounios, J., & Nolde, S. F. (1997). Electrophysiological brain activity and memory source monitoring. Neuroreport, 8(5), 1317-1320.
Johnson, R., Jr., Pfefferbaum, A., & Kopell, B. S. (1985). P300 and long-term memory: latency predicts recognition performance. Psychophysiology, 22(5), 497-507.
Johnston, W. A., Dark, V. J., & Jacoby, L. L. (1985). Perceptual fluency and recognition judgments. Journal of Experimental Psychology: Learning, Memory, and Cognition, 11, 3-11. 130 Karis, D., Fabiani, M., & Donchin, E. (1984). "P300" and memory: individual differences in the von Restorff effect. Cognitive Psychology, 16, 177-216.
Keane, M. M., Gabrieli, J. D., Monti, L. A., Fleischman, D. A., Cantor, J. M., & Noland, J. S. (1997). Intact and impaired conceptual memory processes in amnesia. Neuropsychology, 11(1), 59-69.
Klimesch, W., Doppelmayr, M., Yonelinas, A., Kroll, N. E., Lazzara, M., Rohm, D., et al. (2001). Theta synchronization during episodic retrieval: neural correlates of conscious awareness. Brain Research Cognitive Brain Research, 12(1), 33-38.
Klimesch, W., Hanslmayr, S., Sauseng, P., Gruber, W., Brozinsky, C. J., Kroll, N. E., et al. (2006). Oscillatory EEG correlates of episodic trace decay. Cerebral Cortex, 16(2), 280-290.
Knowlton, B. J., & Squire, L. R. (1995). Remembering and knowing: two different expressions of declarative memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21(3), 699-710.
Kosslyn, S. M., Thompson, W. L., Kim, I. J., & Alpert, N. M. (1995). Topographical representations of mental images in primary visual cortex. Nature, 378(6556), 496-498.
Kutas, M., & Dale, A. M. (1997). Electrical and magnetic readings of mental functions. In M. D. Rugg (Ed.), Cognitive Neuroscience (Studies in Cognition). New York: MIT Press.
Lamme, V. A., & Roelfsema, P. R. (2000). The distinct modes of vision offered by feedforward and recurrent processing. Trends in Neurosciences, 23(11), 571-579.
Leutgeb, J. K., Leutgeb, S., Moser, M. B., & Moser, E. I. (2007). Pattern separation in the dentate gyrus and CA3 of the hippocampus. Science, 315(5814), 961-966.
Levy, D. A., Stark, C. E., & Squire, L. R. (2004). Intact conceptual priming in the absence of declarative memory. Psychological Science, 15(10), 680-686.
Leynes, P. (2002). The effect of test queries on source-monitoring event-related potentials. Brain & Cognition, 50(2), 218-233.
Luce, P. A., & Lyons, E. A. (1998). Specificity of memory representations for spoken words. Memory & Cognition, 26(4), 708-715.
Luck, S. J. (2005). An introduction to the event-related potential technique. Cambridge, Mass.: MIT Press. 131 MacKenzie, G., & Donaldson, D. I. (in press). Dissociating recollection from ‘pure’ familiarity: Electrophysiological evidence from a recognition memory test for faces. Neuroimage, in press.
Mandler, G. (1980). Recognizing: the judgment of previous occurrence. Psychological Review, 87, 252-271.
McCarthy, G., & Wood, C. C. (1985). Scalp distributions of event-related potentials: an ambiguity associated with analysis of variance models. Electroencephalography and Clinical Neurophysiology, 62(3), 203-208.
Mecklinger, A. (2000). Interfacing mind and brain: a neurocognitive model of recognition memory. Psychophysiology, 37(5), 565-582.
Mintzer, M. Z., & Griffiths, R. R. (2000). Acute effects of triazolam on false recognition. Memory & Cognition, 28(8), 1357-1365.
Montaldi, D., Spencer, T. J., Roberts, N., & Mayes, A. R. (2006). The neural system that mediates familiarity memory. Hippocampus, 16(5), 504-520.
Mulligan, N. W. (1998). The role of attention during encoding in implicit and explicit memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 24(1), 27-47.
Mulligan, N. W., & Stone, M. (1999). Attention and conceptual priming: Limits on the effects of divided attention in the category-exemplar production task. Journal of Memory and Language, 41(2), 253-280.
Nessler, D., Mecklinger, A., & Penney, T. B. (2005). Perceptual fluency, semantic familiarity and recognition-related familiarity: an electrophysiological exploration. Brain Research Cognitive Brain Research, 22(2), 265-288.
Nolde, S., Johnson, M. K., & Raye, C. L. (1998). The role of prefrontal cortex during tests of episodic memory. Trends in Cognitive Sciences, 2, 399-406.
Olichney, J. M., Iragui, V. J., Salmon, D. P., Riggins, B. R., Morris, S. K., & Kutas, M. (2006). Absent event-related potential (ERP) word repetition effects in mild Alzheimer's disease. Clinical Neurophysiology, 117(6), 1319-1330.
Olichney, J. M., Van Petten, C., Paller, K. A., Salmon, D. P., Iragui, V. J., & Kutas, M. (2000). Word repetition in amnesia. Electrophysiological measures of impaired and spared memory. Brain, 123(9), 1948-1963. 132 Otten, L. J., Quayle, A. H., Akram, S., Ditewig, T. A., & Rugg, M. D. (2006). Brain activity before an event predicts later recollection. Nature Neuroscience, 9(4), 489-491.
Paller, K. A. (2000). Neural measures of conscious and unconscious memory. Behavioral Neurology, 12(3), 127-141.
Paller, K. A., Bozic, V. S., Ranganath, C., Grabowecky, M., & Yamada, S. (1999). Brain waves following remembered faces index conscious recollection. Brain Research Cognitive Brain Research, 7(4), 519-531.
Paller, K. A., Hutson, C. A., Miller, B. B., & Boehm, S. G. (2003). Neural manifestations of memory with and without awareness. Neuron, 38(3), 507-516.
Paller, K. A., & Kutas, M. (1992). Brain potentials during memory retrieval provide neurophysiological support for the distinction between conscious recollection and priming. Journal of Cognitive Neuroscience, 4, 375-391.
Paller, K. A., Kutas, M., & Mayes, A. R. (1987). Neural correlates of encoding in an incidental learning paradigm. Electroencephalography and Clinical Neurophysiology, 67(4), 360-371.
Paller, K. A., Kutas, M., & McIsaac, H. K. (1995). Monitoring conscious recollection via the electrical activity of the brain. Psychological Science, 6, 107-111.
Paller, K. A., & McCarthy, G. (2002). Field potentials in the human hippocampus during the encoding and recognition of visual stimuli. Hippocampus, 12(3), 415-420.
Paller, K. A., Voss, J. L., & Boehm, S. G. (in press). Validating neural correlates of familiarity. Trends in Cognitive Sciences.
Paller, K. A., & Wagner, A. D. (2002). Observing the transformation of experience into memory. Trends in Cognitive Sciences, 6(2), 93-102.
Penney, T. B., Mecklinger, A., & Nessler, D. (2001). Repetition related ERP effects in a visual object target detection task. Brain Research Cognitive Brain Research, 10(3), 239-250.
Rajaram, S., & Geraci, L. (2000). Conceptual fluency selectively influences knowing. Journal of Experimental Psychology: Learning, Memory, and Cognition, 26(4), 1070-1074.
Ranganath, C., Johnson, M. K., & D'Esposito, M. (2000). Left anterior prefrontal activation increases with demands to recall specific perceptual information. The Journal of Neuroscience, 20(22), RC108. 133 Ranganath, C., & Paller, K. A. (1999). Frontal brain potentials during recognition are modulated by requirements to retrieve perceptual detail. Neuron, 22(3), 605-613.
Ranganath, C., & Paller, K. A. (2000). Neural correlates of memory retrieval and evaluation. Brain Research Cognitive Brain Research, 9(2), 209-222.
Ranganath, C., Yonelinas, A. P., Cohen, M. X., Dy, C. J., Tom, S. M., & D'Esposito, M. (2004). Dissociable correlates of recollection and familiarity within the medial temporal lobes. Neuropsychologia, 42(1), 2-13.
Reber, P. J., Siwiec, R. M., Gitelman, D. R., Parrish, T. B., Mesulam, M. M., & Paller, K. A. (2002). Neural correlates of successful encoding identified using functional magnetic resonance imaging. The Journal of Neuroscience, 22(21), 9541-9548.
Rugg, M. D. (1990). Event-related brain potentials dissociate repetition effects of high- and low- frequency words. Memory & Cognition, 18, 367-379.
Rugg, M. D., & Allan, K. (2000). Event-related potential studies of memory. In E. Tulving & F. I. M. Craik (Eds.), The Oxford Handbook of Memory (pp. 521-537). Oxford: Oxford University Press.
Rugg, M. D., Allan, K., & Birch, C. S. (2000). Electrophysiological evidence for the modulation of retrieval orientation by depth of study processing. Journal of Cognitive Neuroscience, 12(4), 664-678.
Rugg, M. D., & Coles, M. G. H. (1995). Electrophysiology of mind: event-related brain potentials and cognition. Oxford ; New York: Oxford University Press.
Rugg, M. D., & Curran, T. (in press). ERPs and recognition memory. Trends in Cognitive Sciences.
Rugg, M. D., Mark, R. E., Walla, P., Schloerscheidt, A. M., Birch, C. S., & Allan, K. (1998). Dissociation of the neural correlates of implicit and explicit memory. Nature, 392(6676), 595- 598.
Rugg, M. D., & Nagy, M. E. (1989). Event-related potentials and recognition memory for words. Electroencephalography and Clinical Neurophysiology, 72(5), 395-406.
Rugg, M. D., Walla, P., Schloerscheidt, A. M., Fletcher, P. C., Frith, C. D., & Dolan, R. J. (1998). Neural correlates of depth of processing effects on recollection: evidence from brain potentials and positron emission tomography. Experimental Brain Research, 123(1-2), 18-23. 134 Rugg, M. D., & Yonelinas, A. P. (2003). Human recognition memory: a cognitive neuroscience persepective. Trends in Cognitive Sciences, 7(7), 313-319.
Sanquist, T. F., Rohrbaugh, J. W., Syndulko, K., & Lindsley, D. B. (1980). Electrocortical signs of levels of processing: perceptual analysis and recognition memory. Psychophysiology, 17, 568- 576.
Schacter, D. L. (in press). Implicit memory and priming. In J. Byrne (Ed.), Learning and memory: a comprehensive reference. Oxford: Elsevier.
Schacter, D. L., & Buckner, R. L. (1998). Priming and the brain. Neuron, 20(2), 185-195.
Schott, B., Richardson-Klavehn, A., Heinze, H. J., & Duzel, E. (2002). Perceptual priming versus explicit memory: dissociable neural correlates at encoding. Journal of Cognitive Neuroscience, 14(4), 578-592.
Schott, B. H., Richardson-Klavehn, A., Henson, R. N., Becker, C., Heinze, H. J., & Duzel, E. (2006). Neuroanatomical dissociation of encoding processes related to priming and explicit memory. The Journal of Neuroscience, 26(3), 792-800.
Schweinberger, S. R., Pickering, E. C., Burton, A. M., & Kaufmann, J. M. (2002). Human brain potential correlates of repetition priming in face and name recognition. Neuropsychologia, 40(12), 2057-2073.
Sederberg, P. B., Kahana, M. J., Howard, M. W., Donner, E. J., & Madsen, J. R. (2003). Theta and gamma oscillations during encoding predict subsequent recall. The Journal of Neuroscience, 23(34), 10809-10814.
Sederberg, P. B., Schulze-Bonhage, A., Madsen, J. R., Bromfield, E. B., McCarthy, D. C., Brandt, A., et al. (2006). Hippocampal and Neocortical Gamma Oscillations Predict Memory Formation in Humans. Cerebral Cortex.
Senkfor, A. J., & Van Petten, C. (1998). Who said what? An event-related potential investigation of source and item memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 24(4), 1005-1025.
Senkfor, A. J., Van Petten, C., & Kutas, M. (2002). Episodic action memory for real objects: an ERP investigation with perform, watch, and imagine action encoding tasks versus a non-action encoding task. Journal of Cognitive Neuroscience, 14(3), 402-419.
Smith, M. E. (1993). Neurophysiological manifestations of recollective experience during recognition memory judgments. Journal of Cognitive Neuroscience, 5, 1-13. 135 Smith, M. E., & Halgren, E. (1989). Dissociation of recognition memory components following temporal lobe lesions. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15(1), 50-60.
Sperling, R. A., Bates, J. F., Cocchiarella, A. J., Schacter, D. L., Rosen, B. R., & Albert, M. S. (2001). Encoding novel face-name associations: a functional MRI study. Human Brain Mapping, 14(3), 129-139.
Squire, L. R. (2004). Memory systems of the brain: a brief history and current perspective. Neurobiology of Learning and Memory, 82(3), 171-177.
Squire, L. R. (in press). Declarative memory system: retrograde and anterograde amnesia. In J. Byrne (Ed.), Learning and memory: a comprehensive reference. Oxford: Elsevier.
Srinivas, K., & Roediger, H. L. (1990). Classifying implicit memory tests: Category association and anagram solution. Journal of Memory and Language, 29(4), 389-412.
Srinivasan, R., Russell, D. P., Edelman, G. M., & Tononi, G. (1999). Increased synchronization of neuromagnetic responses during conscious perception. The Journal of Neuroscience, 19(13), 5435-5448.
Stark, C. E., & Squire, L. R. (2000). Recognition memory and familiarity judgments in severe amnesia: no evidence for a contribution of repetition priming. Behavioral Neuroscience, 114(3), 459-467.
Thompson-Schill, S. L., D'Esposito, M., & Kan, I. P. (1999). Effects of repetition and competition on activity in left prefrontal cortex during word generation. Neuron, 23(3), 513-522.
Toth, J. P. (1996). Conceptual automaticity in recognition memory: levels-of-processing effects on familiarity. Canadian Journal of Experimental Psychology, 50(1), 123-138.
Trautner, P., Dietl, T., Staedtgen, M., Mecklinger, A., Grunwald, T., Elger, C. E., et al. (2004). Recognition of famous faces in the medial temporal lobe: an invasive ERP study. Neurology, 63(7), 1203-1208.
Trott, C. T., Friedman, D., Ritter, W., Fabiani, M., & Snodgrass, J. G. (1999). Episodic priming and memory for temporal source: event-related potentials reveal age-related differences in prefrontal functioning. Psychology of Aging, 14(3), 390-413.
Tsivilis, D., Otten, L. J., & Rugg, M. D. (2001). Context effects on the neural correlates of recognition memory: an electrophysiological study. Neuron, 31(3), 497-505. 136 Tulving, E. (1985). Memory and consciousness. Canadian Journal of Psychology, 26, 1-12.
Ullsperger, M., Mecklinger, A., & Muller, U. (2000). An electrophysiological test of directed forgetting: the role of retrieval inhibition. Journal of Cognitive Neuroscience, 12(6), 924-940.
Urbach, T. P., & Kutas, M. (2002). The intractability of scaling scalp distributions to infer neuroelectric sources. Psychophysiology, 39(6), 791-808.
Vaidya, C. J., Gabrieli, J. D., Keane, M. M., & Monti, L. A. (1995). Perceptual and conceptual memory processes in global amnesia. Neuropsychology, 9, 580-591.
Van Petten, C., Senkfor, A. J., & Newberg, W. M. (2000). Memory for drawings in locations: spatial source memory and event-related potentials. Psychophysiology, 37(4), 551-564.
Verfaellie, M., & Cermak, L. S. (1999). Perceptual fluency as a cue for recognition judgments in amnesia. Neuropsychology, 13(2), 198-205.
Viskontas, I. V., Knowlton, B. J., Steinmetz, P. N., & Fried, I. (2006). Differences in mnemonic processing by neurons in the human hippocampus and parahippocampal regions. Journal of Cognitive Neuroscience, 18(10), 1654-1662.
Voss, J. L., Baym, C., & Paller, K. A. (in preparation). Implicit memory can support accurate recognition without awareness.
Voss, J. L., Lucas, H., & Paller, K. A. (in preparation). Distinct electrical signals of familiarity and conceptual priming during recognition.
Voss, J. L., & Paller, K. A. (2006). Fluent conceptual processing and explicit memory for faces are electrophysiologically distinct. The Journal of Neuroscience, 26(3), 926-933.
Voss, J. L., & Paller, K. A. (2007). Neural correlates of conceptual implicit memory and their contamination of putative neural correlates of explicit memory. Learning & Memory, 14(4), 259- 267.
Voss, J. L., & Paller, K. A. (in press). Neural substrates of remembering: electroencephalographic studies. In J. Byrne (Ed.), Learning and memory: a comprehensive reference. Oxford: Elsevier.
Voss, J. L., Reber, P. J., Mesulam, M. M., Parrish, T. B., & Paller, K. A. (under review). Familiarity and conceptual priming engage distinct cortical networks. 137 Wagner, A. D., Gabrieli, J. D., & Verfaellie, M. (1997). Dissociations between familiarity processes in explicit recognition and implicit perceptual memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23(2), 305-323.
Wagner, A. D., Koutstaal, W., Maril, A., Schacter, D. L., & Buckner, R. L. (2000). Task-specific repetition priming in left inferior prefrontal cortex. Cerebral Cortex, 10(12), 1176-1184.
Wagner, A. D., Koutstaal, W., & Schacter, D. L. (1999). When encoding yields remembering: insights from event-related neuroimaging. Philosophical Transactions of the Royal Society of London B, 354(1387), 1307-1324.
Wagner, A. D., Shannon, B. J., Kahn, I., & Buckner, R. L. (2005). Parietal lobe contributions to episodic memory retrieval. Trends in Cognitive Sciences, 9(9), 445-453.
Wais, P. E., Wixted, J. T., Hopkins, R. O., & Squire, L. R. (2006). The hippocampus supports both the recollection and the familiarity components of recognition memory. Neuron, 49(3), 459- 466.
Wheeler, M. E., & Buckner, R. L. (2004). Functional-anatomic correlates of remembering and knowing. Neuroimage, 21(4), 1337-1349.
Wheeler, M. E., Shulman, G. L., Buckner, R. L., Miezin, F. M., Velanova, K., & Petersen, S. E. (2006). Evidence for separate perceptual reactivation and search processes during remembering. Cerebral Cortex, 16(7), 949-959.
White, N. (in press). Multiple parallel memory systems in the brain. In J. Byrne (Ed.), Learning and memory: a comprehensive reference. Oxford: Elsevier.
Whittlesea, B. W., & Williams, L. D. (1998). Why do strangers feel familiar, but friends don't? A discrepancy-attribution account of feelings of familiarity. Acta Psychologia (Amsterdam), 98(2- 3), 141-165.
Wilding, E. L. (2000). In what way does the parietal ERP old/new effect index recollection? International Journal of Psychophysiology, 35(1), 81-87.
Wilding, E. L. (2006). On the practice of rescaling scalp-recorded electrophysiological data. Biological Psychology, 72, 325-332.
Wilding, E. L., & Rugg, M. D. (1996). An event-related potential study of recognition memory with and without retrieval of source. Brain, 119, 889-905. 138 Winkelman, P., Halberstadt, J., Fazendeiro, T., & Catty, S. (2006). Prototypes are attractive because they are easy on the mind. Psychological Science, 17, 799-806.
Wixted, J. T. (2007). Dual-process theory and signal-detection theory of recognition memory. Psychological Review, 114(1), 152-176.
Wixted, J. T. (in press). Dual-process theory and signal-detection theory of recognition memory. Psychological Review.
Wixted, J. T., & Squire, L. R. (2004). Recall and recognition are equally impaired in patients with selective hippocampal damage. Cognitive, affective & behavioral neuroscience, 4(1), 58-66.
Wixted, J. T., & Stretch, V. (2004). In defense of the signal detection interpretation of remember/know judgments. Psychonomic Bulletin & Review, 11(4), 616-641.
Woldorff, M. G., Liotti, M., Seabolt, M., Busse, L., Lancaster, J. L., & Fox, P. T. (2002). The temporal dynamics of the effects in occipital cortex of visual-spatial selective attention. Brain Research Cognitive Brain Research, 15(1), 1-15.
Wolk, D. A., Schacter, D. L., Berman, A. R., Holcomb, P. J., Daffner, K. R., & Budson, A. E. (2005). Patients with mild Alzheimer's disease attribute conceptual fluency to prior experience. Neuropsychologia, 43(11), 1662-1672.
Woodruff, C. C., Hayama, H. R., & Rugg, M. D. (2006). Electrophysiological dissociation of the neural correlates of recollection and familiarity. Brain Research, 1100(1), 125-135.
Yonelinas, A. P. (2001). Consciousness, control, and confidence: the 3 Cs of recognition memory. Journal of Experimental Psychology: General, 130(3), 361-379.
Yonelinas, A. P. (2002). The nature of recollection and familiarity: A review of 30 years of research. Journal of Memory and Language, 46(3), 441-517.
Yonelinas, A. P., Hopfinger, J. B., Buonocore, M. H., Kroll, N. E., & Baynes, K. (2001). Hippocampal, parahippocampal and occipital-temporal contributions to associative and item recognition memory: an fMRI study. Neuroreport, 12(2), 359-363.
Yonelinas, A. P., Kroll, N. E., Dobbins, I., Lazzara, M., & Knight, R. T. (1998). Recollection and familiarity deficits in amnesia: convergence of remember-know, process dissociation, and receiver operating characteristic data. Neuropsychology, 12(3), 323-339. 139 Yonelinas, A. P., Kroll, N. E., Quamme, J. R., Lazzara, M. M., Sauve, M. J., Widaman, K. F., et al. (2002). Effects of extensive temporal lobe damage or mild hypoxia on recollection and familiarity. Nature Neuroscience, 5(11), 1236-1241.
Yonelinas, A. P., Otten, L. J., Shaw, K. N., & Rugg, M. D. (2005). Separating the brain regions involved in recollection and familiarity in recognition memory. The Journal of Neuroscience, 25(11), 3002-3008.
Yovel, G., & Paller, K. A. (2004). The neural basis of the butcher-on-the-bus phenomenon: when a face seems familiar but is not remembered. Neuroimage, 21(2), 789-800.
140 Table 1 A memory taxonomy based on findings in amnesic patients.
Type of memory Behavioral outcome Findings in patients with amnesia Declarative Recall and recognition of episodes and Impaired storage, causing memory facts (i.e., episodic memory and semantic deficits in new learning and in memory) remembering information acquired prior to onset of amnesia Immediate Information available while kept in mind Preserved if performance is not memory by continuous rehearsal (e.g., verbal supported in part by retrieving working memory) declarative memories for the rehearsed information Nondeclarative memory (a large category that includes Generally preserved, but with nonassociative learning, classical conditioning, category some notable exceptions learning, habit learning, as well as the following) Perceptual Speeded or more accurate responses in Preserved if performance is not priming a priming test, based on item-specific contaminated by declarative or perceptual representations memory
Conceptual Speeded or more accurate responses in Preserved in some cases, but priming a priming test, based on association- further investigation is required, specific or conceptual representations particularly across stimulus domains Skills Behaviors that improve gradually Preserved when skill acquisition with practice, including cognitive is accomplished without reliance skills (e.g., reading mirror-reversed on declarative memory (i.e., not text) and motor skills most skills learned outside laboratory circumstances)
141 Table 2 fMRI results for famous faces. Summary of regional activation clusters for each experimental contrast, including Broadman’s area (BA), hemisphere, centroid Talairach-Tournoux coordinates, volume, and mean statistical difference across the cluster.
Experimental Contrast Centroid coordinates Volume Side 3 t value Region (BA) x y z (mm )
Conceptual Priming: Primed minus Unprimed Superior frontal gyrus (8) L -16 38 48 391 -3.7 Inferior frontal gyrus (45/47) L -50 19 3 203 -3.5 Fusiform/parahippocampal gyrus (19/37) L -27 -56 -6 422 3.6 Middle temporal gyrus (39/19) L -44 -74 16 1000 3.9 Posterior cingulate gyrus (31) L -8 -63 16 609 3.9 Anterior cingulate gyrus (24) L 3 21 9 234 3.9
Nonspecific Explicit Memory: HEM minus LEM Inferior/superior parietal lobule (40/7) R 43 -42 47 2527 4.2 Angular gyrus (39) L -43 -70 27 297 4.0 Medial frontal gyrus (10) L -1 62 10 859 4.7 Posterior cingulate gyrus (31) L -6 -56 23 516 3.8 Inferior temporal gyrus (19) L -52 -72 -1 469 4.1
Episodic Familiarity: Know minus New Inferior/superior parietal lobule (40/7) R 35 -41 56 4446 4.7 Middle occipital gyrus (19) L -29 -63 3 967 -4.6 Inferior frontal gyrus (9) L -44 10 29 250 4.0 Inferior frontal gyrus (9) R 25 20 -12 344 -4.3 Superior frontal gyrus (6) L -6 7 51 781 4.4 Lingual gyrus (18) L -16 -73 -8 422 -4.0 Cingulate gyrus (18) L 1 -28 27 656 3.9
Nonspecific Explicit Memory minus Conceptual Priming Superior frontal gyrus (8) L -15 29 51 500 3.8 Inferior frontal gyrus (45/47) L -52 30 4 266 3.7 Inferior parietal lobule (40) R 43 -40 48 1996 4.1 Paracentral lobule (5) L -6 -39 57 578 4.0
Episodic Familiarity minus Conceptual Priming Superior frontal gyrus (8) L -18 33 51 1185 3.8 Inferior frontal gyrus (45/47) L -51 18 3 500 3.6 Inferior/superior parietal lobule (40/7) R 35 -39 57 4774 4.3 Putamen R 28 2 2 686 3.9 Lingual gyrus (19) L -26 -63 -1 344 -4.0 Medial frontal gyrus (10) R 14 38 -10 641 -3.9 Subcallosal gyrus (47) R 20 19 -10 484 -3.9
142 Table 3
Conceptual priming for squiggles of high subjective meaningfulness.
(a) Meaningfulness Rating Test for (b) Meaningfulness Rating Test with (c) Loop Discrimination Test for Perceptual Conceptual Priming (RT) Repetition Acknowledged (RT) Priming (Proportion Correct)
High-M Low-M New High-M Low-M New High-M Low-M New
788 (34)* 836 (48) 834 (39) 1007 (59)** 968 (56)** 846 (31) 0.95 (1.3)** 0.94 (1.2)** 0.89 (0.9)
Numbers in parentheses indicate SE of the mean. Significant priming effects relative to a baseline from new items are indicated (* = p<0.05 and ** = p<0.01).
143 Table 4
Recognition performance for uncommon words. Calculation of d′ for each meaningfulness category utilized a common new item baseline. Meaningful and Meaningless items given high- confidence know responses, the two categories of old items that contribute to ERP contrasts, are
highlighted in green and yellow, respectively. Missed items assigned to the miss category are
highlighted in gray.
High- Medium- Low- Remember Confidence Confidence Confidence New Know Know Know
New 1.1 (0.5) 4.7 (1.1) 6.8 (1.1) 20.4 (2.8) 67.2 (3.3) Percent new stimuli
Old Meaningful 20.1 (2.2)** 19.3 (2.4) 5.8 (0.9) 2.6 (0.5)* 1.2 (0.3)* Percent repeated stimuli 1.3 (0.1)** 0.8 (0.1) -0.1 (0.1) -1.0 (0.1)** Discrimination sensitivity (d′)
Old Meaningless 8.5 (1.6)** 16.0 (2.1) 7.2 (0.8) 5.9 (1.1)* 2.9 (0.5)* Percent repeated stimuli 0.7 (0.1)** 0.7 (0.1) 0.1 (0.1) -0.7 (0.1)** Discrimination sensitivity (d′) * p<0.05 for meaningful vs. meaningless pariwise difference within a single response category. ** p<0.001 for meaningful vs. meaningless pariwise difference within a single response category. 144 Table 5
Conceptual priming for uncommon words.
Meaningful Old Meaningful New Meaningless Old Meaningless New
Lexical Decision Accuracy (%) 83.1 (6.1) 85.9 (5.5) 82.1 (2.9) 84.5 (4.8) Reaction Time (ms) 690.6 (32.4)* 767.4 (42.5)* 756.0 (39.4) 731.1 (40.3)
Liking Rating Reaction Time (ms) 982.5 (48.6) 945.1 (36.7) 1038.7 (41.2) 978.6 (50.8) Rating 2.62 (0.06)* 2.47 (0.05)* 2.24 (0.05) 2.27 (0.09) Accuracy in the lexical decision task is the percentage of uncommon words (which subjects did not know were real words) that were correctly identified as nonwords. *Pairwise old/new difference for the given meaningfulness category p<0.05.
145 Table 6
Formal comparisons of ERP correlates of recognition memory. ERP correlates of recognition memory were obtained by comparing repeated stimuli correctly endorsed with high-confidence familiarity responses to missed repeated stimuli with factors: Condition (meaningful and miss or meaningless and miss), Region (seven regions, identified in Figure 1), and Latency Interval (300-
500, 500-700, and 700-900 ms). ANOVA results and the mean correct/miss ERP difference amplitudes averaged over each region and latency interval identified via a main effect or interaction involving the Condition factor are indicated.
Latency Interval (ms) 300-500 500-700 700-900 Meaningful vs. Miss Condition-by-Region F(3.2,54.6)=2.3, p=0.08 Mid-frontal 0.9* 1.1 0.7 Main effect F(1,17)=3.6, p=0.08 Right-frontal 0.9* 0.6 0.5 Main effect F(1,17)=4.4, p=0.05 Central 0.8 1.0* 0.9* Main effect F(1,17)=5.6, p=0.03 Left-posterior 0.7 1.5* 1.4* Main effect F(1,17)=7.1, p=0.02 Mid-posterior 1.2 1.4* 1.3* Main effect F(1,17)=7.0, p=0.02 Right-posterior 0.7 1.1* 1.0* Main effect F(1,17)=4.7, p=0.05 Meaningless vs. Miss 3-way Interaction F(3.2,54.0)=5.32, p=0.002 Central -0.3 1.5* 0.4 Interaction F(1.7,28.3)=4.7, p=0.02 Left-posterior 0.7 2.4* 1.4 Interaction F(1.6,26.8)=5.3, p=0.01 Mid-posterior 1.4 2.5* 1.0 Interaction F(1.6,26.4)=4.8, p=0.02 Right-posterior 0.2 1.7* 0.4 Interaction F(1.5,24.7)=3.1, p=0.06 * Pairwise old/miss difference significant following Bonferroni correction (p<0.017).
146 Figure 1
Visualizing ERPs. Waveforms averaged across experimental subjects for two experimental conditions, A and B, are considered at a single electrode (top). These waveforms are plotted with time shown on the x-axis going from left to right, and with voltage on the y-axis. Here, positive potentials are plotted in the upward direction, although some investigators plot amplitudes in the opposite manner. The time course of the difference in amplitude between these conditions can be visualized as a difference wave (middle), or by averaging over latency intervals of interest (gray shading) for many scalp locations and generating images of the voltage differences across the scalp (bottom). These topographic maps schematically represent the distribution of ERP differences between conditions A and B for the two latency intervals. The head is viewed from above, approximated by a circular shape, with anterior scalp regions towards the top. Black dots indicate the locations of recording electrodes. The black circle identifies the electrode location used for the waveforms shown. These images can thus demonstrate both temporal and spatial characteristics of an ERP effect. Reproduced from Voss and Paller (in press). 147
148 Figure 2
Schematic of the Dm or subsequent-memory methodology. During an encoding phase, neuroimaging measures are recorded while subjects attempt to remember stimuli. During the subsequent retrieval phase, memory tests for these stimuli are administered. Performance on these tests is used to classify study-phase stimuli and corresponding neural measures into two categories, subsequently-remembered and subsequently-forgotten. A comparison between neural correlates of these two conditions thus yields neurophysiological differences computed on the basis of subsequent memory performance, Dm. Figure adapted from Paller and Wagner (2002) and reproduced from Voss and Paller (in press).
149 Figure 3
Representative Dm effects for two categories of memory. Two different memory experiences can be assessed after people learn to associate novel faces with randomly assigned occupations.
Yovel and Paller (2004) categorized trials according to whether faces were forgotten or remembered with reference to the face alone with no contextual retrieval (familiarity) or remembered with retrieval of both the face and the paired occupation (recollection). ERPs were recorded when subjects studied these face/occupation pairs, and relatively more positive ERPs were found to predict later memory. These Dm effects are displayed at two representative electrodes in A and the scalp topography of these effects appears in B. Scalp maps are of the head viewed from above, with anterior oriented toward the top. Comparing the two effects, Dm for recollection exhibited a bilateral topography with larger amplitudes spanning a longer time interval, whereas Dm for familiarity was smaller, briefer, and restricted to right posterior locations. Figure adapted from Yovel and Paller (2004).
150 151 Figure 4
Late positive potentials index conscious recollection. Posterior, positive potentials from 300 to
800 ms after the presentation of faces that were previously encoded with the instruction to remember (Remember faces) were greater than for faces that were encoded with the instruction to forget (Forget faces). Each Remember face was also presented with a short biographical vignette, and subjects were highly accurate at recollecting this information when cued by the corresponding face. Behavioral results also showed that recognition was better for Remember faces than for Forget faces whereas priming did not differ. Thus, these late positive potentials index recollective processing uncontaminated by priming. Data are from Paller et al. (1999).
Reproduced from Voss and Paller (in press).
152 Figure 5
Late frontal potentials associated with post-retrieval processing. ERPs to drawings of common objects were compared between a highly demanding (specific) recognition test and a less-demanding (general) recognition test. Waveforms (left) showed a relative positivity for the specific test compared to the general test for all three stimulus classes in the experiment: objects that are perceptually identical to one in the study phase (old/same), objects that are perceptually altered from one in the study phase (old/different), and entirely novel object pictures. The corresponding ERP topography was computed over all conditions for the latency interval from
500-1200 ms. These late frontal potentials were maximal over the left anterior scalp, but in other experiments have also been found to be bilaterally symmetric or maximal over right anterior scalp. Figure adapted from Ranganath and Paller (1999) and reproduced from Voss and Paller (in press).
153 Figure 6
ERP correlates of recollection and perceptual priming. The ERP difference between
recollected faces and new faces is displayed in A. The difference between primed but not
remembered faces and new faces is displayed in B. Differences are averaged over 100 ms intervals starting at the latency indicated underneath each topographic map. Light colors indicate positive difference potentials in A and negative difference potentials in B. Figure adapted from
Paller et al. (2003) and reproduced from Voss and Paller (in press).
154 Figure 7
Schematic representation of the experimental design in Experiment 1, including timing
parameters. The construction of Phase 1 produced two classes of famous faces, primed and
unprimed, that varied systematically in the degree to which associated conceptual knowledge
was activated. The biographical information applied only to the primed faces (e.g., the three
biographical cues depicted here apply to the primed celebrity). Brain potentials were recorded in
Phase 2 while speeded responses to each famous face were obtained. Deciding that a face is
famous tends to entail access to relevant biographical facts, and we hypothesized that this
response would be facilitated for faces to the extent that associated biographical information was
activated in Phase 1. Primed and unprimed faces were presumably subject to equivalent perceptual priming from viewing faces in Phase 1. Conceptual priming was thus exhibited in
Phase 2 by faster and more accurate responses for primed compared to unprimed faces. An assessment of explicit memory in Phase 3 revealed the extent to which each celebrity was known to each participant. Brain potential differences based on conceptual priming and explicit memory
were thus contrasted for the same set of stimuli within the same task. Reproduced from Voss and
Paller (2006). 155
156 Figure 8
ERPs elicited by famous faces during the Priming Test. a ERPs for primed and unprimed conditions (N=10) and HEM and LEM conditions (N=8) at frontal (slightly posterior to Fz) and parietal (Pz) midline locations. Significant pairwise differences for 5-ms windows are indicated by black bars below each axis. The assignment of famous faces to primed and unprimed conditions was counterbalanced across subjects, such that reliable differences cannot be attributed to the specific faces used in each condition. b Topographic maps of mean ERP differences between primed vs. unprimed and HEM vs. LEM conditions averaged over two time intervals. The three electrode regions for formal analyses are indicated in the Explicit Memory
250-500 ms map: frontal and posterior regions by large dots (16 and 18 electrodes, respectively), middle region by small dots (18 electrodes). c ERPs from the same two locations as in (a) for the primed/unprimed contrast for HEM faces only (N=8) and for the HEM/LEM contrast for unprimed faces only (N=8). d Topographic maps of mean ERP differences for the corresponding two contrasts from (c) averaged over two time intervals. Reproduced from Voss and Paller
(2006). 157
158 Figure 9
ERPs elicited by nonfamous faces to produce neural correlates of pure familiarity, shown in the same format as in Fig. 2. Results were reported in detail by Yovel and Paller (2004). a ERPs for faces recognized with pure familiarity and new faces at electrode locations matching those displayed in Fig. 2 (top: Fz, bottom: Pz). b Topographic maps of mean ERP differences based on the ERPs in (a) and averaged over two time intervals. These results show that neural correlates of pure familiarity bear a strong similarity to neural correlates of explicit memory for famous faces
(Fig. 2, right side). Reproduced from Voss and Paller (2006).
159
Figure 10
Schematic representation of the behavioral paradigm. The four experimental phases are
depicted, showing the contrasts used to isolate neural correlates of three mnemonic phenomena
— conceptual priming, nonspecific explicit memory, and episodic familiarity.
160 Figure 11
Double dissociation between fMRI correlates of conceptual priming and familiarity. a
Three regions (black voxels) were identified via double subtractions between conceptual priming and nonspecific explicit memory and between conceptual priming and episodic familiarity.
Regions of left superior frontal gyrus (SFG, BA 8) and inferior frontal gyrus (IFG, BA 45/47) were also identified by the conceptual priming contrast. The region of right inferior parietal lobule was also identified by both the nonspecific explicit memory contrast and the episodic familiarity contrast (IPL, BA 40, centroid TT coordinates=43 x, -42 y, 49 z, volume=1203 mm3). b The average magnitude of the BOLD response from 5-9 s for each experimental condition is shown averaged over these regions. c The double dissociation is apparent in the average differences between conditions that constituted the three experimental contrasts. Error bars indicate SE. The reported effects reflect signal enhancements or reductions in positive-going
BOLD responses (estimated impulse response functions for each condition appear in Figure 12). 161
162
Figure 12
Estimated impulse response functions for each condition. Functions were averaged across subjects and within each region identified in Figure 11 (IFG, top row; SFG, middle row; IPL, bottom row). Error bars indicate across-subject SE.
163 Figure 13
Examples of squiggle stimuli. a A very small number of squiggle stimuli were given high and low meaningfulness ratings by a majority (>80%) of subjects (upper and lower two rows, respectively). b The majority of stimuli, such as these, were given inconsistent ratings across subjects. (Stimuli were from Groh-Bordin et al., 2006.) Reproduced from Voss and Paller
(2007).
164 Figure 14
Study- and test-phase behavioral results in Experiment 3. a Mean proportion of old and new squiggles given remember (R), know (K), guess (G), or new (N) responses during test phase in
Experiment 2. b Mean proportion of squiggles (200 total) endorsed with each meaningfulness rating level during study phase. c Mean proportion of old squiggles given remember, know, guess, and new responses during test phase, subdivided by study-phase meaningfulness ratings
(High-M or Low-M). Error bars indicate SE. d Across-subjects meaningfulness rating histogram for every squiggle (300 total), sorted by consistency. Light grays indicate that a low proportion of subjects endorsed the squiggle with the corresponding rating, whereas dark grays indicate a high proportion of consistent ratings (mean rating σ=1.46). Reproduced from Voss and Paller
(2007).
165 Figure 15
ERP correlates of episodic memory. a Anterior, middle, and posterior scalp regions, as seen on a schematic view of the head shown from above. ERPs in this and subsequent figures were computed by spatially averaging responses over these three regions (A, anterior, white; M, middle, light gray; P, posterior, dark gray). b Waveforms to correctly-identified old squiggles of both meaningfulness levels (All-M) and to correctly-rejected new squiggles averaged spatially by region. c Topographic maps of the All-M old vs. new ERP difference, averaged over three latency intervals (highlighted by shading on ERP waveforms: 300-500 ms, 500-700 ms, and 700-
900 ms). Reproduced from Voss and Paller (2007).
166
Figure 16
ERP correlates of episodic recollection and familiarity. a Spatially-averaged waveforms (see
Fig. 3a) to All-M old squiggles correctly identified with remember and know responses, and to
correctly-rejected new squiggles. b Topographic maps of the old vs. new ERP difference
averaged over two latency intervals separately for remember (top) and know (bottom) responses.
ERPs to remember responses were much larger than to know responses, and so the color scales for amplitude were set differently to allow the topographic patterns of the two effects to be observed clearly. Reproduced from Voss and Paller (2007).
167 Figure 17
ERP correlates of conceptual priming. a Spatially-averaged (see Fig. 3a) waveforms to
squiggles given know responses made separately for High-M and Low-M conditions. b
Topographic maps of the conceptual priming ERP difference (High-M-know minus Low-M-
know) averaged over two latency intervals. c Mean High-M-know and Low-M-know ERP amplitudes at each latency interval and region. Stars mark statistically significant differences.
Error bars indicate SE after correcting for between-subject variability in mean values over all conditions. Reproduced from Voss and Paller (2007).
168 Figure 18
Correlations between behavioral estimates of familiarity-based recognition and ERPs.
Across subjects, d' for High-M-Know responses (black squares and lines) and Low-M-Know
responses (gray squares and lines) correlated significantly with the corresponding ERP old/new
effect for middle (top) and posterior (bottom) scalp regions from 500-700 ms. No correlations
involving the anterior region or 300-500 ms latency interval reached statistical significance.
Reproduced from Voss and Paller (2007).
169
Figure 19
Identification of electrodes within each scalp region. A single electrode within each of the
seven scalp regions is identified according to the 10-20 naming convention. The lower case
letters following electrode labels indicate that the given electrode was slightly anterior (a),
posterior (p), superior (s), or inferior (i) to the corresponding 10-20 electrode. The regions
included the approximate 10-20 electrodes: left-frontal: F3a, F7a, F3i, F7p, C5a, C3a; mid-
frontal: Fza, F3s, F4s, Fzp, FC1, FC2, Cza; right-frontal: F4a, F8a, F4i, F8p, C6a, C4a; central:
C1a, C2a, Cz, C1p, C2p, Pzs; left-posterior: C3’, P3a, P1’, P3i; mid-posterior: Pzi, PO1, PO2,
Ozs, O1’, O2’, O1i, O2i, Ozi; right-posterior: C4’, P4a, P2’, P4a.
170 Figure 20
Neural correlates of conceptual priming. ERP results are shown for meaningful and meaningless words restricted to those recognized with high-confidence know responses.
Waveforms in the upper right correspond to the three electrodes indicated with red circles on the
schematic head diagram in the upper left (small black dots indicate the positions of electrodes on
the scalp as viewed from above, with the nose at the top). Neural correlates of conceptual
priming (meaningful minus meaningless difference for high-confidence know responses) are
shown topographically for three latency intervals (bottom). Difference values less than 0 V
were nonsignificant and are shown in black.
171 Figure 21
Neural correlates of recognition memory. a ERPs to meaningful and meaningless words, restricted to those recognized with high-confidence know responses, are juxtaposed to ERPs to old words that were missed. b Topographic plot of ERP differences between meaningful high- confidence know words and miss words (recognition with concomitant conceptual priming), averaged over three latency intervals. c Topographic plot of ERP differences between meaningless high-confidence know words and miss words (recognition without concomitant conceptual priming), averaged over three latency intervals.
172 Appendix 1 Celebrity biographical cues organized by set. Set 1 Name Cue 1 Cue 2 Angelina Jolie Laura Croft Tomb Raider Gone in 60 Seconds Aretha Franklin Sang 'Respect' Queen of Motown Barbara Streisand Sang 'You Don't Buy Me Flowers' Prince of Tides Brandy Sang 'That Boy is Mine' Show of pregnancy Calista Flockhart Allie McBeal A Midsummer Night’s Dream Carmen Electra Till Death Do Us Part Scarry Movie Catherine Zeta Jones Chicago Traffic Cher Sang 'Believe' Sang 'And the Beat Goes On' Christina Aguilera Sang 'Dirty' Sang 'Beautiful' Christina Ricci Sleepy Hollow Monster Claudia Schiffer Supermodel Zoolander Demi Moore G.I. Jane Strip Tease Elizabeth Hurley Austin Powers Estee Lauder spokeswoman Goldie Hawn The Banger Sisters First Wives’ Club Halle Berry Monster’s Ball Bond Girl Helen Hunt As Good as it Gets Mad About You Jamie Lee Curtis Freaky Friday Halloween movies Janeane Garofalo Comedian and actress Mystery Men Janet Reno 1st female Attorney General Parkinson's disease Jennifer Love Hewitt Sang 'Bare Naked' Party of Five Jenny Jones Daytime talk show hostess Show taped in Chicago Joan Lunden Good Morning America Spokesperson for Vaseline Intensive Care Julia Louis Dreyfus Elaine on Seinfeld Watching Ellie sitcom Juliette Lewis Starsky & Hutch Natural Born Killers Katie Holmes Joey on Dawson’s Creek Phone Booth Kelly Rowland Destiny's Child album 'Simply Deep' Kirstie Alley Look Who’s Talking mother Cheers Lida Kudrow Phoebe on Friends Analyze This Lucy Liu Charlie’s Angels Kill Bill Madonna Sang 'Like a Virgin' Sang 'Like a Prayer' Marla Shriver First Lady of California Ex-news reporter Michelle Pfeiffer I am Sam White Oleander Mira Sorvino Summer of Sam Mighty Aphrodite Missy Elliot Sang 'Pass the Dutch' Sang 'One Minute Man' Nicole Kidman The Hours Moulin Rouge Oprah Winfrey Hosts Oprah Wealthiest woman in America Pink Sang 'You make me sick' Sang 'Just like a pill' Queen Latifah Chicago Bringing Down the House Renee Zellweger Chicago Bridget Jones’s Diary Roseanne Arnold Roseanne Bad national anthem Sandra Bullock Miss Congeniality Forces of Nature Shannon Doherty Hosts Scare Tactics Beverly Hills 90210 Reba McEntire country music singer Reba sitcom Rosie O’Donnel Rosie talk show Recent same-sex marriage Sarah Ferguson Former Duchess of York Spokesperson for Weight-Watchers Tyra Banks Victoria’s Secret model Halloween: Resurrection Adam Sandler Billy Madison 50 First Dates Alex Trebek Jeopardy Host Gameshow host Emmy nominee Albert Einstein Patent Clerk Relativity Theory Antonio Banderas Spy Kids father Once Upon a Time in Mexico Ben Stiller Meet the Parents Starsky and Hutch Bill Maher Politically Incorrect Political activist Bob Costas Sportscaster Internight host Christopher Reeves Superman Paraplegic Conan O'Brien Late Show Comic insult dog Dan Rather CBS evening news 60 Minutes II David Arquette Eight-Legged Freaks Scream movies David Letterman Late Night Stupid human tricks David Spade Joe Dirt Tommy Boy Dustin Hoffman Rainman The Graduate Emilio Esteves Mighty Ducks Young Guns 173 Name Cue 1 Cue 2 Eminem 8 Mile Sang 'Lose Yourself' Harrison Ford Indiana Jones Han Solo Harry Caray Sportscaster Restaurant chain Jason Alexander George on Seinfeld Shallow Hal Jay Leno Late Night Classic car collector Jesse Jackson Campaigned for presidency Religious political activist John Goodman Roseanne husband Oh brother, where art thou? John Lennon Sang 'Imagine' shot in 1980 John Travolta Swordfish Saturday Night Fever Joshua Jackson Pacey on Dawson's Creek Cruel Intentions Kelsey Grammer Frasier Cheers Leonardo Dicaprio Gangs of New York Titanic LL Cool J album 'Phenomenon' album 'Walking with the Panther' Marilyn Manson Bowling for Columbine Sang 'Beautiful People' Matt Damon Rounders Saving Private Ryan Matthew Broderick Ferris Beuler Inspector Gadget Michael Jackson Sang 'Thriller' Jackson Five Michael J. Fox Back to the Future Teenwolf Norm McDondald Saturday Night Live Dirty Work Notorious BIG Biggie Smalls died in 1997 Peter Jennings World News Tonight The Century co-author Prince William Prince of Wales Lives in Windsor Palace Richard Gere Chicago Primal Fear Richard Nixon Resigned from Presidency President during Vietnam War Roger Ebert Film Critic Sun Times columnist Robin Williams Insomnia Death to Smoochy Rudy Giuliani Former mayor of NYC Times 2001 person of the year Sean Connery Original James Bond The Rock Shaquille O’Neal Basketball player Shazam
Set 2 Name Cue 1 Cue 2 Barbara Walters 20/20 The View Beyonce Knowles Sang 'Crazy in Love' Destiny's Child Britney Spears Sang 'Hit me Baby One More Time' Sang 'Oops, I did it again' Celine Dion Sang 'My heart will go on' Married her older manager Christina Applegate Married with Children The Sweetest Thing Chyna Playboy Female professional wrestler Cindy Crawford 90's model Pepsi spokesperson Courney Cox Monica on Friends Scream movies Courtney Love Hole lead singer Married Kurt Cobain Ellen Degeneres Finding Nemo Ellen sitcom Gwen Stefani No Doubt lead singer Sang 'I'm just a girl' Gwenyth Paltrow Shakespeare in Love Shallow Hal Hillary Clinton Former First Lady Senator of New York Jane Pauley Dateline NBC Longtime Today anchor Janet Jackson Sang 'All for you' Clothing malfunction Jennifer Aniston Rachel on Friends Office Space Jennifer Lopez Sang 'Jenny from the Block' The Wedding Planner Jewel Sang 'Who will save your soul?' Sang 'Foolish Games' Jodie Foster Silence of the Lambs Panic Room Judge Judy TV Judge Worked in NY family court Julia Roberts Pretty Woman Mona Lisa Smile Kate Winslet Titanic Eternal Sunshine of the Spotless Mind Katie Couric Today Show Colon cancer spokesperson Kirsten Dunst Spiderman Bring It On Lara Flynn Boyle The Practice Ballerina Dress at Golden Globes Liv Tyler Armageddon Lord of the Rings Maria Carey Sang 'Butterfly' Sang 'Heartbreaker' Maya Angelou Sang 'I know why the caged bird sings' Presidential inaugural poem Melissa Joan Hart Sabrina the Teenage Witch Clarissa Explains it All Minnie Driver Circle of Friends Good Will Hunting Monica Lewinsky White House intern Presidential sex scandal 174 Name Cue 1 Cue 2 Neve Campbell Party of Five Scream movies Paula Abdul American Idol Sang 'Straight Up' Penelope Cruz Vanilla Sky BLOW Princess Diana Princess of Wales Died in car crash Reese Witherspoon Legally Blonde Cruel Intentions Ricki Lake Hairspray Daytime talk show hostess Rose McGowan Jawbreaker Charmed Sally Fields Forest Gump The Flying Nun Sally Jesse Raphael Daytime talk show hostess Red glasses Selma Hayek Frida Once Upon a Time in Mexico Sharon Stone Basic Instinct Casino Sheryl Crow Sang 'All I wanna do' Sang 'Strong Enough' Suzanne Sommers Three's Company Thighmaster Uma Thurman Pulp Fiction Kill Bill Al Gore Former Vice President Lost 2002 Presidential election Bill Clinton Former governor of Arkansas Presidential sex scandal Bob Barker The Price is Right Happy Gilmore Charles Barkley Basketball "Dream Team" NBA studio commentator Colin Powell Secretary of State 4-star General U.S. Army Cuba Gooding Jr. Jerry McGuire Radio Dana Carvey CNL Church Lady Wayne's World David Schwimmer Ross on Friends The Pallbearer Drew Carey The Drew Carey Show Whose line is it anyway? Ed Norton Fight Club American History X Elton John Sang 'Candle in the Wind' Sang 'Tiny Dancer' George Clooney E.R. Doctor Ocean's 11 George W. Bush President of U.S. War on Terror Hugh Downs 20/20 ABC news team anchor James Van Der Beek Dawson on Dawson's Creek Varsity Blues Jerry Springer Former mayor of Cincinnati Trashy talk show JFK Jr. George magazine Died in private plane crash Jim Carrey Dumb and Dumber Ace Ventura Johnny Depp Pirates of the Caribbean BLOW Josh Hartnett Pear Harbor 40 Days and 40 Nights Justin Timberlake N'Sync Sang 'Justified' Keanu Reeves The Matrix Bill and Ted's Excellent Adventure Kevin Bacon Footloose Mystic River Mahatma Gandhi Passive resistance Fasted to quell riots Marv Albert Sports announcer Sex scandal Matt LeBlanc Joey on Friends Charlie's Angels Matthew Perry Chandler on Friends Fools Rush In Mel Gibson The Passion Braveheart Michael Jordan Chicago Bulls number 13 Hanes Underware ads Mike Meyers Wayne's World Austin Powers OJ Simpson Murder defendant White Bronco Pierce Brosnan Newest James Bond The Thomas Crowne Affair Prince Charles Prince of Wales Affair with Camilla Puff Daddy P. Diddy Bad Boy Records Ricki Martin Sang 'Livin' La Vida Loca' Sang 'She Bangs' Ronald Regan President and actor Iran Contra scandal Robert Redford Horse Whisperer Indecent Proposal Russell Crowe Gladiator Master and Commander Saddam Hussein Former Iraqi President Invaded Kuwait Seth Green The Italian Job Austin Powers Simon Cowell American Idol Record producer Ted Koppel Nightline anchor Nighline managing editor Tobey McGuire Spiderman Seabiscuit Tom Hanks Forest Gump Apollo 13 Val Kilmer The Saint Batman Forever
175 Appendix 2
List of the uncommon words used as stimuli. abacinate brogan dryad leal prow abscind burke eclat lepid pugilist abstruse cache egress levant purloin accipiter cafard empasm limn putcher accubation calamus epicure limous quarion acoria calver esker littoral quirt aculeate camber facula ludic rachis acumen camorra falderal mabble ragmatical addle canard fanal macrural raiment adenia cang fane maffick ranarian adze canthus fardage manus rebus afferent carnifex farthing marl recreant agger catena ferity mascaron remora ague cavil fipple matinal rhumb alacrity cenoby foin maunder rillet alar chaton forel menald robur algid choller fossor metis rosin allodic cimex frith mica ruelle almuce cladose fulgent minaret sagacity ament claver gavage monad sarment amity cloaca gemel monture scelerat anneal coffle gingham morass sellate anthelion coif glozing morkin sepiment antiphon columella goral mucin sessile argol comate gricer mulm sideral artifice comestible gurlet nadir skelder asperity conatus gybe nascent solander assibilate copacetic haslock nepenthe strepent aval corf henotic olamic strop avaunt cornice hopple onager sural avowal cortege incanous operculum swage baft crampon incult osculation sward bannock crebrous indurate pabulum tain bavin crotal inumbrate paean tapis bechic culet jamb palfrey tarn beloid curtate jerid palinola tephra benthic cylix jib pangram thane betel dalliance jorum parnel tonsure bilious dapatical jupon patulous trammel birl darby kale pavis trellis blissom debel kedge pergola unction blunge dedans knoll phaeton urman blype delf kurgan pinguid varec bodge desudation lacis pizzle vestal brayer diadem lagan plangent vitriol bream didact lambent praxis wimple bricole dight larkspur prosaic witan brio drail latrant provender yapness