Sternberg verbal working memory task

The experimental task was a variant of the Sternberg task (Sternberg 1966). In each trial, subjects had to keep a briefly presented set of consonants in memory by rehearsal until a probe letter was presented, and to then indicate by button press whether this probe matched one of letters in the initially presented set. The Sternberg task can be divided into discrete phases of stimulus encoding, maintenance, and retrieval which differ in terms of sensory and motor demands. The present variant of the task used normal and degraded stimulus quality (see Fig. 1) to tax sensory acquisition demand (please see below for behavioral findings from a supplementary study that confirmed increased sensory acquisition demand for the present degraded compared to the non-degraded stimuli). To tax oculomotor demand, stimulus spacing was varied in two levels (close or wide). Luminance was equated for all stimuli. To manipulate cognitive and articulatory load, two and four letter consonant sets were used. As shown in Fig. 1, percent sign (%) placeholders were displayed in unused positions in the consonant sets with only two letters. As mentioned in the previous study (Peterburs et al. 2016), the task also included overt and covert rehearsal instructions at the beginning of each trial. However, since sensory acquisition was expected to only affect the encoding phase, trials were pooled across rehearsal conditions for the present analyses. In addition, trials were also pooled across stimulus spacing conditions.

Schematic illustration of the sequence and time course of stimulus presentation in a trial of the Sternberg verbal working memory task

Figure 1 illustrates the sequence and time course of stimulus presentation in one trial of the Sternberg task. Each trial started with a fixation cross that was presented for 3–5 s (which implemented jitter in trial onsets for fMRI analysis), followed by a set of two or four consonants that was presented for 2 s (encoding phase). A blank screen was presented in the maintenance phase for 4 or 6 s, while subjects were rehearsing the consonants. Last, the probe was presented for 3 s, and subjects decided whether it was a “match” or “no match” by pressing one of two response buttons with their right index or middle finger. Task instructions emphasized both speed and accuracy, and accordingly, response time (RT) and accuracy were recorded for each trial. Subjects completed 10 practice trials outside the scanner prior to starting the experiment. The task consisted of three runs with 64 trials each, amounting to 192 trials in total. Stimulus quality (degraded or normal), spacing (wide or close), and cognitive load (high or low) were balanced in each run. Trial order within each run was pseudorandomized so that presentation of identical parameters was limited to three consecutive trials. Each run contained 24 unique match and 24 unique no-match trials as well as 16 trials without a probe. The latter were included to allow the hemodynamic response to fully return to baseline following the maintenance phase. In these trials, a blank screen was presented in the retrieval phase, and no response was expected. For trials with probe, positions of the target letter in the initial sequences were counterbalanced. Task completion took approximately 55 min and included short breaks between runs.

Stimulus presentation was controlled with E-Prime 2 software (Psychology Software Tools Inc., Sharpsburg, PA, USA) running on a Hewlett Packard xw4300 workstation. The visual display was rear-projected onto a screen in the MRI scanner located behind the participant and then reflected onto a mirror fixed to the head coil within the participant’s immediate line of view. Responses were recorded using two Velcro-connected fiber optic button boxes (MRA, Inc., Washington, PA).