We present the generalized signal detection theory (GSDT), where familiarity is described by a sparse binomial distribution of binary node activity rather than by normal distribution of familiarity. Items are presented in a distributed representation, where each node receives either noise only, or signal and noise. An old response (i.e., a "yes" response) is made if at least one node receives signal plus noise that is larger than the activation threshold, and item variability is determined by the distribution of activated nodes as the threshold is varied. A distinct representation leads to better performance and a lower ratio of new to old item variability, than a more distributed and less distinct representations. Here we apply the GSDT to empirical data on verbal and olfactory memory and suggest that verbal memory relies on a distinct neural item representation, whereas olfactory memory has a fuzzy neural representation leading to poorer memory and inducing a larger ratio of new to old item variability.

Signal detection theory (SDT) and the Dual Process SDT (Yonelinas, 1994) are currently the most influential accounts of item variability in recognition memory. However, neither provides a sufficient account of differences in the familiarity distributions. Instead, this phenomenon is accounted for by the idea of encoding variability (Wixted, 2007) or an additional retrieval process (Yonelinas, 2001). We present the Generalized Signal Detection Theory (the GSDT), in which the familiarity distribution are a sum of signals described by a sigmoidal non-linear activation function. The GSDT accounts for a higher variability in the old item distribution by emphasizing the non-linarites, but also for equal variability in the new and old item distributions by attenuating the non-linearites. The GSDT also extends the interpretation of the new to old item variability, indexed by the slope of the z-ROC. (Less)

Stenberg et al. argued that FN400 brain potentials index familiarity rather than conceptual priming. Their data from a test of name recognition showed that both familiarity and FN400s were influenced by frequency but not fame, whereas separate behavioral measures of priming were influenced by fame but not frequency. However, this apparent dissociation was gravely weakened by confounds in task demands and inadequate behavioral measures of priming. Although Stenberg et al. failed to provide evidence suitable for disentangling neural correlates of familiarity from those of conceptual priming, an analysis of their report can be used to highlight difficulties that remain to be surmounted to understand recognition and the neural events that signal distinct memory functions engaged during recognition.

■ Lucas, Voss, and Paller sympathize with our intentions but disagree with our findings. They argue that a relation between frequency and conceptual priming may have been obscured by methodological details in our second experiment, therefore failing to complete a bridge between conceptual priming and FN400 with name frequency as the mediator. However, renewed inspections of our experiment and a new additional experiment, designed to meet the objections, fail to find any role for name frequency in conceptual priming and therefore re-establish the dissociation of priming and the FN400. On closer inspection, our differing views seem to derive from different interpretations of the term “concept.” ■ Despite the vast number of ERP studies of recognition memory, there have been only a few studies demonstrating a double dissociation of the putative ERP correlates of recollection and familiarity within an experiment ( Jäger, Mecklinger, &Kipp, 2006;Woodruff, Hayama,&Rugg, 2006). Our recent article (Stenberg, Hellman, Johansson, & Rosén, 2009) joins those findings by showing that orthogonal manipulations of frequency and fame selectively influence contributions of familiarity (FN400 old/new effect) and recollection (left parietal old/new effect) to recognition memory for names. Strong support for dual-process accounts of recognition memory can be derived from these data. We further argued that our data can inform interpretations of the FN400 old/new effect, a currently debated issue. In a second behavioral experiment, fame and not frequency influenced measures of conceptual priming. Given the strong relationship between the frequency and the FN400 old/new effect, we concluded that this early old/ new effect seemed to index familiarity (cf. Rugg & Curran, 2007) rather than conceptual priming (cf. Paller, Voss, & Boehm, 2007). Although Lucas, Voss, and Paller (2010) profess sympathy for our attempts to disentangle familiarity and conceptual priming, they take issue with our conclusions, against which they present three main points of criticism. They argue that a relation between frequency and conceptual priming (a) may have been overlooked in our analysis and (b) was not given a fair chance to develop. Further, they argue that (c) the relation between frequency and familiarity, which we did find, was not the one that the familiarity hypothesis predicted. SENSITIVITY OF THE CONCEPTUAL PRIMING MEASURE Lucas et al. (2010) argue that an analysis of a subset of the data from Experiment 2, that is, the fame decision task, would provide a more sensitive test of an effect of frequency on conceptual priming. We ran the requested analysis and failed anew to find any reliable influence of frequency, F(1,12) = 1.98, ns. In terms of effect sizes, the effect of fame was five times stronger than that of frequency. To support their position, Lucas et al. would have to explain not only that frequency, while producing the FN400, failed to give conceptual priming but also that fame, while giving rise to much stronger conceptual priming, failed to produce the FN400. COMPARABILITY OF CONCEPTUAL PROCESSING ACROSS EXPERIMENTS Lucas et al. (2010) note that the encoding tasks were different and suggest that the intentional encoding in Experiment 1 served to enhance conceptual fluency for infrequent names and incidental encoding, with a frequency-decision orienting task, in one condition of Experiment 2, favored famous names. From their comments, it remains unclear why a frequency-decision task should prime famous names selectively and not frequent/infrequent names. Wouldnʼt it be much simpler to assume, as we do, that the fame dimension makes a real difference for conceptual priming because people have concepts associated with famous people, and concepts are invariably found at the business end of conceptual priming? Frequency, on the other hand, does not make a difference because rare, unheard of, even outlandishly quaint names do not carry conceptual content. Theymay trigger a stray, idiosyncratic association, yes, and in this diluted sense theymay prime something, but for substantial priming to take place the association will have to be reliably repeated. Even if one were to accept the assumptions of Lucas et al. (2010) regarding conceptual processing for minimally meaningful stimuli and, for the sake of argument, © 2009 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 22:4, pp. 618–620 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.2009.21268 by guest on 04 June 2021 concede that such processing occurred for rare names in Experiment 1, a lot of explaining would still remain to be done. If nonfamous names could release such conceptual processing, how much more wouldnʼt we expect from famous names? The requirements for conceptual processing are fulfilled, a host of facts are known about these persons that can be summoned up at will. The stage would be set for substantial, reliably repeated conceptual processing, which, according to the view of Lucas et al., should be accompanied by a large FN400 effect. Yet, we see none; why? On one point, we grant that Lucas et al. (2010) have pinpointed a shortcoming in our study. The encoding conditions in the two experiments were not identical, and through this loophole, noise may have entered. Because it is not possible to argue about this possibility by armchair theorizing alone, we ran a follow-up behavioral conceptual priming study to amend and assess the consequences. The experiment, comprising 24 participants, used study instructions identical to those in Experiment 1 (i.e., “remember as much as you can”) and a test phase identical to the speeded fame judgment in Experiment 2. Thus, the conditions implemented the very conjunction of encoding task and priming measurement, to which Lucas et al. (2010) appealed. With a type of encoding that gives free rein to idiosyncratic associations, opportunities are maximal for the kind of processing that is included in the generous definition of conceptual priming (Paller et al., 2007). With a priming task that has a proven track record (Experiment 2), measurement should be sensitive enough to detect all possible traces of conceptual priming associated with frequency. We ran the experiment (with Inquisit, on the Internet, just as Experiment 2), and the results are as follows: The effect of celebrity on RT priming scores (new minus old) was highly reliable, F(1,23) = 57. 36, p < .001, and the effect size (partial η) was large (0.71). The effect of frequency, on the other hand, was nonsignificant F(1,23) = 1.36, p > .25, and negligible, η = 0.06. There was no interaction (Figures 1 and 2). In all relevant aspects, this follow-up experiment reproduced the outcome of Experiment 2. Fame sustained priming, and frequency did not. Thus, our efforts to disclose any hitherto hidden links between frequency and conceptual priming, and by extension any link between conceptual priming and the FN400, failed even under these favorable conditions. The encoding condition was the one designated by Lucas et al. (2010) as likely to promote conceptual processing for all names, whether famous or not. FINDINGS “CONTRADICT AN ASSOCIATION BETWEEN FN400 AND FAMILIARITY” Lucas et al. (2010) argue that our findings of a strong coupling of FN400 old/new effect and rare names contradict an association between this ERP effect and familiarity. By this they mean that high-frequency names should be expected to be familiar and therefore produce an FN400 effect, which they did not. After all, what could be more familiar than names such as Smith and Jones? True, but not quite to the point. As we argue in the article, quoting Mandlerʼs (1980) seminal work, it is the familiarity increment that counts, that is, the increase a studied object incurs, going from its resting level to a freshly activated state. This step can be huge for a rare and odd name; for a common name, it may be negligible. In addition, as predicted by this hypothesis, familiarity in the recognition test was indeed much higher for rare names than for common names. Common names had clearly lower familiarity estimates in the behavioral data (see dp values in Table 1 of the original article) than rare names did. Lucas et al. (2010) insist that common names evoked “high levels of familiarity.” If they are referring to an absolute benchmark level above which a dp value may be called high, then this gold standard is unknown to us. Figure 1. RTs for the old and new names in the priming task. Name types are formed by the independent combination of frequency and celebrity in high (+) or low (−) degrees. Note that reactions to famous names grow both faster and more accurate as they are primed fame. Figure 2. Error rates for the old and new names in the priming task. Note that the phenomenon of “false fame”—falsely designating as famous a name that has been recently primed—applies to some extent to common, nonfamous names but not to rare, nonfamous names. Therefore, the lack of RT priming for this group cannot be ascribed to hesitancy and equivocation caused by false fame.

Recent interest has been drawn to the separate components of recognition memory, as studied by event-related potentials (ERPs). In ERPs, recollection is usually accompanied by a late, parietal positive deflection. An earlier, frontal component has been suggested to be a counterpart, accompanying recognition by familiarity. However, this component, the FN400, has alternatively been suggested to reflect a form of implicit memory, conceptual priming. The present study examined the ERP components of recognition memory using an episodic memory task with a stimulus material consisting of names, half of which were famous. Along a different dimension, the names varied in how rare or common they were. These dimensions, frequency and fame, exerted powerful effects on memory accuracy, and dissociated the two recognition processes, such that frequency gave rise to familiarity and fame fostered recollection, when the receiver operating characteristics data were analyzed with Yonelinas' dual-process signal detection model. The ERPs corresponded fully to the behavioral data because frequency affected the frontal component exclusively, and fame affected the parietal component exclusively. Moreover, a separate behavioral experiment showed that conceptual priming was sensitive to fame, but not to frequency. Our data therefore indicate that the FN400 varies jointly with familiarity, but independently of conceptual priming.

Pre-experimental familiarity can have paradoxical effects on episodic memory Knowledge of the stimulus domain usually enhances memory, but word frequency-a presumed correlate of prior experience-is negatively related to recognition accuracy The present study examined episodic recognition of names and its relation to two measures of pre-experimental knowledge, name frequency, and fame. Frequency was operationalised as the number of hits in a national telephone directory, and fame as hits on national mass media websites. Recognition accuracy was increased by fame, but diminished by frequency. Four experiments confirmed the findings, using yes/no recognition, ROC curves, and remember-know paradigms. Hit rates were consistently more strongly influenced by fame than by frequency, whereas the reverse was true for false alarm rates. These dissociations suggest that two different forms of semantic memory, specific and nonspecific knowledge, interact with episodic memory in separate ways.