Aim 1.1: Causal role of readiness potentials in conscious intention.

In this experiment, subjects view a self-paced slide show with the slide being advanced when an EEG brain-computer interface (BCI) detects that the subject is about to act, or by the subject pressing a button, whichever comes first. After each slide advance the subject reports the degree to which s/he felt responsible for the slide advance on a continuous scale from “certain I did it” to “certain the computer did it”. Normally a BCI would be trained to detect whatever features in the signal are most predictive of the to-be-detected event (a button press). But in this case the BCI would be trained to specifically detect RP-like events in the EEG data, even if these are less predictive than other signals that might be present. We hypothesize that the amplitude of the averaged readiness potential will be greater for slide advances when the subject feels that s/he had changed the slide, compared to when the subject feels that the computer changed the slide. Follow-up studies will focus on identifying cortical sources that have the strongest association with the feeling of conscious will and testing the causal role of those sources (using TMS).

Aim 1.2: Causation between intention and action representations.

As outlined above, a different question about the causal role of intentions pertains to their consequences, that is whether they are causes of the subsequent action. Here we will address this issue using neural signals that provide a more immediate handle on the processes involved. The idea is thus to assess the causal dependency between brain representations of intentions and subsequent actions, thus testing the hypothesis that intentions are not merely epiphenomenal. For this, it is necessary to identify cases where the representations of intentions and actions can be clearly anatomically and temporally separated in the brain, which is difficult for motor-intentions and readiness potentials. Thus, we will choose a different cognitive domain and only employ motor intentions as a control. We will engage a variety of different kinds of intentions within the same experiment in order to be able to identify the regions, processes, pattern of activity, and patterns of functional connectivity that encode intentions in a more abstract and general form across different kinds of intentions. This focus on abstract stages of intention representation will allow us to separate between brain signals pertaining to intentions and actions.

Aim 2.1: Unconsciously producing preferential bias in arbitrary intentions.

In this project, we plan to test the effect of consciously and unconsciously presented primes on subjects’ decisions. The latter might either be arbitrary or deliberate decisions. Priming will be measured in two ways; first, probability of making a decision that is congruent with the prime, compared with a decision that is incongruent with the prime. Second, reaction time when making a decision that is congruent with the prime, compared with a decision that is incongruent with the prime (see Figure 4, taken from the grant). To maximize the chances of obtaining an effect, we will focus on difficult decisions, where the difference in value between the options is relatively small (this will be piloted, to make sure the task is not too frustrating for subjects).

Aim 2.2: Interaction between causation and consciousness.

Subjective reports of intention onset (Libet et al., 1983; Matsuhashi & Hallett, 2008) vary dramatically in the onset of conscious intentions that they find. Yet, a key aspect of an intention is that it settles behavior. Such settledness also seems to appear when decoding action contents over time from brain activity (see plateau in Fig.—from Salvaris & Haggard, 2014). We will therefore test the relation between the level of conscious, subjective settledness of a decision—as it is being made—and the amount of information in the brain about that decision—as decoded by a classifier from brain activity. The existence of such a relation would suggest that the onset of a plateau in the accuracy of a neural classifier could be used as an objective measure of intention onset.

We will run a version of the experiment in Study 1.2 adapted for EEG—with a pre-cue baseline period, shorter delay period, and shorter time to respond after the go signal. The highly variable delay period and rapid response required motivates participants to form an intention by the start of the delay period. This will comprise 75% of the trials. The other 25% will be catch trials using a version of the probe method (Matsuhashi & Hallett, 2008). A tone will be presented to the subject at a random time during the delay period, and they will be instructed to report the strength of their belief that they had already formed an intention at probe onset on a Likert scale from 1 (certain of no intention) to 7 (certain of intention). We hypothesize that the rise in subjects’ certainty about their intentions will track the rise of the accuracy of our decoder.

Aim 2.3: The causal import of attention.

The goal of Aim 2.3 is to investigate the causal role of covert attention in decision-making. The experiment specifically examines whether voluntarily directed, covert attention makes a difference to behavior. In the experiment, subjects will decide whether or not to act in three different kinds of situations: (1) when the act promises the possibility of reward for the subject; (2) when the act risks harm to another person; and (3) when the act has both features—the subject has a chance of benefiting and, also, a third party has a chance of being harmed. On each trial, the kind of situation is signaled by colored disks appearing to the left and right of a fixation cross, and covert attention is directed to one or the other of the two disks2. We predict that for trials with both reward and loss, attending to the reward or loss will increase or decrease the empirical probability that subjects will decide to act, respectively.

Aim 2.4: Does becoming aware of unconscious events increase volitional control over those events?

This aim was not included in the final version of the grant. It was discussed in the meeting, and we decided to try and think about it some more, to see if we can flesh it out and design a good-enough experiment to test it. Thus, currently this is the most under-developed aim in this project.

The main thrust is to manipulate subjects’ awareness of inner, neural events of which the subject is unaware, and measure the effect this manipulation has on behavior. This will be done via real-time neural feedback that will specifically target an unconscious process of interest (on which we need to decide; this could be the process leading to an upcoming perceptual alternation in binocular rivalry, or an increased response in the amygdala towards a black face.

These are mere examples, that are not obligating in any way). Generally speaking, we ask if making subjects aware of an unconscious event allows them to exert volitional control over it. Notably, making them aware of this state entails also providing them with information about it; we will need to see if the two can be disentangled, and if not – decide if this can still have implications on the question of consciousness’ role in volitional control. A possible idea is to try and give the feedback itself subliminally, so that the system is presented with the information in both cases, once with the awareness of it and once without it.

Aim 2.5: Reaction-time experiments comparing responses to conscious and unconscious stimuli (using masking or TMS).

Using the same stimulus and the same response with a consciously seen stimulus and a masked stimulus, it should be possible to compare, at least in this special circumstance, conscious vs. unconscious control of movement. With backward masking, for example, persons do not see the masked stimulus, but still react to it. The experiment can be done with simple or choice reaction time. In the simple reaction time situation, persons believe they are responding to the mask, where they are really responding to the unseen stimulus. Studies can also be done without masking; then they both see and respond to the stimulus. In the choice reaction time situation, the first stimulus and the mask, can indicate different movements. In this circumstance, when responding in the masked situation, it will appear to the subject as a mistake.

Aim 3.1: Compare the neural correlates of arbitrary and deliberate intentions.

A key EEG correlate of upcoming actions, the readiness potential (RP), is known to precede subjects’ reports of their decision to move. Some view this as evidence against a causal role for consciousness in human decision-making and thus against free-will. Yet those studies focused on arbitrary decisions—purposeless, unreasoned, and without consequences. It remained less clear to what degree the RP generalizes to deliberate, more ecological decisions. A previous study directly compared deliberate and arbitrary decision-making during a $1000-donation task to non-profit organizations. They found the expected RPs for arbitrary decisions. But those were strikingly absent for deliberate ones. Those results and their accompanying drift-diffusion model were congruent with the RP representing the accumulation of noisy, random fluctuations that drive arbitrary—but not deliberate—decisions. They further point to different neural mechanisms underlying deliberate and arbitrary decisions, challenging the generalizability of studies that argue for no causal role for consciousness in decision-making to real-life decisions.

In the current aim, we focus on the neural precursors of deliberate decisions, as these were not investigated in the above study. Based on previous research, those precursors might be in brain regions that are not accessible using EEG. In addition, there is evidence from fMRI studies for neural precursors of arbitrary decisions. In this sub-aim, we will therefore directly compare the neural precursors of arbitrary and deliberate decisions using fMRI.

Aim 3.2: Does perturbing the neural correlates of arbitrary intentions affect deliberate intentions?

The notion that the readiness potential (RP) is a ballistic neural activation, with its onset matching the onset of the decision—or even reflecting the decision-making process—has been challenged in recent years (e.g., Schurger et al., 2012). What is more, it has been demonstrated that the RP might be limited to arbitrary decisions (as in Libet et al., 1983) and not generalize to deliberate ones (Maoz et al., under review). Here we will investigate the degree to which the RP plays a causal role in arbitrary and in deliberate decisions. We will dampen the RP using rTMS and then investigate any differential behavioral affect that it would have on arbitrary and deliberate decisions, if at all.

Aim 3.3: Do subjects have less access to their arbitrary intentions than to their deliberate ones?

TMS pulses to the primary motor cortex, which reliably caused reflex-like contralateral index-finger flexion, were triggered off the EMG activity of the voluntary, arbitrary flexion of that index finger (see Fig.). The TMS click was after the beginning of the muscle contraction and hence happened after the onset of volition. However, subjects reported being on the verge of deciding when hearing the TMS click and witnessing their finger move, hence while their will was still unsettled. Any decent definition of intention would suggest it has occurred at TMS onset. Yet, delivering a TMS pulse that would trigger an action that has already begun tends to result in an erroneous inference about intention onset. Here we will compare the effect of the above for arbitrary and deliberate decisions.

There are a number of concepts that are central to accounts in the neuroscience of intentions and free will, such as “intention”, “consciousness”, “representation”, “decision”, “volition”, “freedom”, and of course “free will”. Many of these concepts are contested: they are understood in different ways by different theorists, and are often understood in different ways within different disciplines. The aim of this aspect of the project is to identify some of the key differences here, and to ask foundational questions about how these concepts should be understood.