Figure 1.
In Signal Detection Theory (SDT), studied (old) items elicit a higher feeling of familiarity relative to unstudied (new) items, both with a normal distribution.
The mean familiarity difference between the old and new item distributions reflects recognition memory (old/new discrimination performance, or d′). By contrast, the point on the familiarity dimension above which “old” responses are rendered and below which “new” responses are rendered reflects the decision criterion (or response bias, c). More items with relatively low familiarity are contained in the Hit response category (blue area) when the threshold is liberal as compared to conservative; the same is true for Correct Rejections (red area). In the Two High Threshold Model (2 HTM), Hit Rates (HR) reflect the probability that an old item is either correctly recognized (Pr), or not (1 - Pr) while there is a bias to respond “old” (Br). This is mathematically formulated as: [1] HR = Pr+(1 - Pr) Br. On the other hand, False Alarm Rates (FAR) reflect the probability that a new item is not recognized (1 - Pr) while there is a bias to respond “old” (Br), mathematically formulated as: [2] FAR = (1 - Pr) Br. Using [2] to resolve Pr in [1] yields: HR = Pr+FAR, equivalent to Pr = HR – FAR. Rewriting [2] yields Br = FAR/(1 - Pr). Since recognition thresholds for old and new items are assumed to be equal, Pr and Br can be computed analogously from Correct Rejection Rates and Miss Rates. Thick lines in the decision tree depict different response tendencies under conditions of relatively liberal (high Br) as opposed to conservative (low Br) decision criteria. For both models, SDT and 2 HT, if accurate old/new discrimination is equal, changes in Hits, Correct Rejections, or any response types, can only stem from bias, whether these are changes in response rates, confidence ratings, or associated brain measures.
Figure 2.
Hitrates (HR) and False Alarm Rates (FAR, top), Bias Br and Accurate Old/New Recognition Pr (center), and RTs (bottom) for Experiment 1 (error bars display standard errors).
Figure 3.
Only data of the 24 participants who varied the decision criterion in line with instructions are included in this Figure.
TOP: Frontocentral grand average ERPs for the random condition (left) and the block condition (right) of Experiment 1. The negative component (FN320) peaking around 320 ms was larger (more negative) for the liberal than for the conservative decision criterion. BOTTOM: Parietal grand average ERPs for the random condition (left) and the block condition (right) of Experiment 1. The displayed waveforms are averaged across the seven sites included in the statistical analysis (see inserts).
Table 1.
Results of the ANOVA of frontal ERP amplitudes.
Table 2.
Results of the ANOVA of parietal ERP amplitudes.
Figure 4.
Mean ERP amplitudes taken in early (300–500 ms poststimulus) and late (500–700 ms poststimulus) time-windows at frontal (AFz, AF3, AF4, F1, F2, FCz, Fz) and parietal (see insert in Figure 3) sites in Experiment 1.
Figure 5.
Hitrates (HR) and False Alarm Rates (FAR, top), Bias Br and Accurate Old/New Recognition Pr (center), and RTs (bottom) for Experiment 2 (error bars display standard errors).
Note that Br and Pr are computed from the same False Alarm Rates for both levels of familiarity since familiarity was not manipulated in new items in this design (but see Experiment 3).
Figure 6.
Frontocentral grand average ERPs for old (left) and new (right) items of Experiment 2.
The negative FN320 component peaking around 320 ms was larger (more negative) for the liberal than for the conservative decision criterion. BOTTOM: Parietal grand average ERPs for the random condition (left) and the block condition (right) of Experiment 2. The displayed waveforms are averaged across the seven sites included in the statistical analysis (see inserts).
Figure 7.
Mean ERP amplitudes taken in early (300–500 ms poststimulus) and late (500–700 ms poststimulus) time-windows at frontal (Fz, AFz, AF3, AF4, F1, F2, FCz) and parietal (see insert in Figure 6) sites in Experiment 2.
Figure 8.
Hitrates (HR) and False Alarm Rates (FAR, top), Bias Br and Accurate Old/New Recognition Pr (center), and RTs (bottom) for Experiment 3 (error bars display standard errors).
Figure 9.
Frontocentral (site FCz) grand average ERPs for high familiarity items (left) and the low familiarity items (right) of Experiment 3.
The FN320 was larger (more negative) for the liberal than for the conservative decision criterion only in the low familiarity condition. In the later time window around 600 ms poststimulus, the ERP old/new differences were larger in the conservative condition than in the liberal condition. BOTTOM: Parietal grand average ERPs for the low familiarity (left) and the high familiarity condition (right) of Experiment 3. The parietal waveform was averaged across the seven sites included in the statistical analysis (see insert).
Figure 10.
Mean ERP amplitudes taken in early (300–500 ms poststimulus) and late (500–700 ms poststimulus) time-windows at frontal (Fz, AFz, AF3, AF4, F1, F2, FCz) and parietal (see insert in Figure 9) sites in Experiment 3.