What I Say is What I Get: Stronger Effects of Self-Generated vs. Cue-Induced Expectations in Event-Related Potentials

Abstract

Expectations regarding future events enable preparatory processes and allow for faster responses to expected stimuli compared to unexpected stimuli. Expectations can have internal sources or follow external cues. While many studies on expectation effects use some form of cueing, a direct comparison with self-generated expectations involving behavioral and psychophysiological measures is lacking. In the present study we compare cue-induced expectations with self-generated expectations that are both expressed verbally in a within-subjects design, measuring behavioral performance, and event-related brain potentials (ERPs). Response time benefits for expected stimuli are much larger when expectations are self-generated as compared to externally cued. Increased amplitudes in both the N2 and P3 components for violations of self-generated expectations suggest that this advantage can at least partially be ascribed to greater perceptual preparation. This goes along with a missing benefit for stimuli matching the expected response only and is mirrored in the lateralized readiness potential (LRP). Taken together, behavioral and ERP findings indicate that self-generated expectations lead to increased premotoric preparation compared to cue-induced expectations. Underlying cognitive or neuronal functional differences between these types of expectation remain a subject for future studies.

Keywords: N2; P3; cue-induced expectations; event-related brain potentials; lateralized readiness potential; self-generated expectations.

Figure 1

Task used in the experiment. On each trial, participants either had to read aloud the cue (in this case “star”) or to verbalize their prediction for the upcoming stimulus (in this case “yellow”). After 1,000 ms the fixation point appeared on screen and after another 1,500 ms the stimulus appeared on screen (in this case red in the prediction trial, signifying a mismatch, and a star in the cued trial, signifying a match). The participants manually responded to the stimulus by pressing one of two keys. The next trial started 500 ms after the response. The mapping of the stimulus types (shape, color) to the expectation conditions and the order of the expectation conditions was balanced over participants.

Figure 2

Response times and error rates. Response times (outer bars) exhibit an effect of match (with faster responses to stimulus matches than to response matches and mismatches) and an interaction with expectation condition (with a stronger effect of match in the self-generated predictions, left). The same pattern is visible in the error rates (inner bars). There were less errors made in the match trials, thus the effect in response times cannot be explained by a speed-accuracy trade-off. Error bars represent confidence intervals (95%) for repeated-measures designs according to Loftus and Masson (1994) and Jarmasz and Hollands (2009).

Figure 3

ERPs at midline electrodes Fz, Cz, and Pz time-locked to stimulus onset. Prediction condition is shown on the left, cue condition on the right. For each electrode the different waveforms for the three match types are shown. Analysis windows for N2 and P3 components are highlighted in gray. Stimulus matches are marked by the solid line, response matches by the dashed line, and mismatches by the dotted line. The interaction of match and expectation condition can best be seen at the Fz electrode for the N2 and at the Pz electrode for the P3.

Figure 4

Top: stimulus-locked LRP waveforms for the prediction condition (left) and cue condition (right). There was an earlier S-LRP onset for stimulus matches than for mismatches and response matches. (Onsets are marked by the short vertical lines intersecting the waveforms.) This onset difference was, in trend, larger for the prediction condition. Although the response match S-LRP onset is as late as for mismatches, they differ in their amplitude before S-LRP onset (50% of the maximum amplitude) in the time interval 150–250 ms following stimulus onset (highlighted in gray). The response match amplitude rises in the correct direction as with the stimulus match and is significantly higher than the mismatch amplitude, but only in the prediction condition. Bottom: response-locked LRP waveforms for the prediction condition (left) and cue condition (right). There is only a significant effect of match with an earlier LRP-R onset for stimulus matches compared to response matches and mismatches.