Human motivation theories are supposed to explain the energization and direction of behavior. In their target article, Murayama and Jach (M&J) argue that theories involving broad constructs such as goals, motives, and needs are “pitched too high” for this purpose. Specifically, they assert that traditional motivational concepts, like intrinsic motivation (IM) or needs for competence, autonomy, or belongingness, play no role in energizing behavior. What really causes behavior is some array of simpler cognitive mechanisms as addressed by the authors’ reinforcement learning paradigm (Figure 2). Understanding motivation is thus not about understanding people, as they attempt to discern and meet their own needs; rather, it is about understanding cognitive processes that are largely inscrutable to the people that they run.

These are bold conclusions, given that decades of research have established the explanatory and practical utility of higher-order motivation constructs. For example, psychological autonomy (i.e., feeling ownership of one's behavior), measured in different ways and contexts, has emerged as critical for persistent engagement and mental health (Ryan et al., Reference Ryan, Duineveld, Di Domenico, Ryan, Steward and Bradshaw2022). In education, meta-analyses show that interventions boosting teachers’ autonomy support reliably enhance student IM, engagement, and performance (e.g., Reeve, Ryan, Cheon, Matos, & Kaplan, Reference Reeve, Ryan, Cheon, Matos and Kaplan2022). Yet such social- and personality-level factors do not count as causes from these authors’ computational perspective.

So, what motivational energizers are identified by their reinforcement learning paradigm? Scrutiny of Figure 2 reveals some puzzles in this regard. In that figure, first note that all downstream processes are driven (starting at the top of the figure) by “awareness of a knowledge gap.” But doesn't such awareness imply a pre-existing motivation to know or learn something? If the person has no such desire, they will not perceive or care about the gap. Notably, IM (wanting to do an activity because it is interesting) is known to enhance people's sensitivity to knowledge gaps.

Befitting the article's reductionistic stance, however, no motives are represented in Figure 2. Still, arrows lead from “existing knowledge network” (bottom) to “the generation of new questions” (middle right), to “awareness of a knowledge gap” (top). This sequence appears to assume that people want to increase their knowledge of the world (a hallmark of IM theories), but this is not explained. Instead, we are informed that “In the reward-learning framework, there is an implicit assumption that people choose to seek information that has a high reward value.” This is the circular problem most reinforcement theories have had, as they are devoid of content regarding experiential value and reward. In the center-right of Figure 2 one finds “rewarding experiences,” but these energize information-seeking behavior only via a side-loop affecting “the expected reward value of new information,” which is said to moderate the path from knowledge gap to information-seeking. In short, M&J's model replaces a relatively straightforward scheme (a person pursues knowledge in domains in which they have interest or value, and that motivation can be enhanced or diminished by experienced supports and obstacles) with a less intuitive scheme, in which the energizers of behavior are either unspecified, or split up amongst an array of low-level process variables.

Why do we need higher-order motivation constructs? There are many possible justifications. One is that cybernetic/hierarchical models of action control assert that much behavior is energized and directed by the abstract or long-term goals a person adopts (Carver & Scheier, Reference Carver and Scheier1981). For example, the goal “I will become a researcher!” has likely organized the daily behaviors of many BBS readers. Similarly, broad motive dispositions (e.g., nAchievement) are known to result from childhood environments and affordances (McClelland, Koestner, & Weinberger, Reference McClelland, Koestner and Weinberger1989) that set parameters for what people strive for throughout their lives (Sheldon & Schuler, Reference Sheldon and Schuler2011). As a third example, self-determination theory (Ryan & Deci, Reference Ryan and Deci2017) shows that people better internalize and sustain motivation for activities in which psychological needs for competence, relatedness, and autonomy can be satisfied, predicting learning and engagement over time.

In short, we would argue that causality is not invariantly bottom-up as M&J imply. Instead, the “low level mental computations” they highlight may better be thought of as part of the how of motivation (i.e., the ways in which our preferences and motives are executed), not the why of motivation (i.e., its energizers). They are mechanisms which serve our varied goals and motives, rather than always determining them.

Indeed, Figure 1B contains a very relevant arrow, which M&J don't discuss, that leads down from “subjective experiences” to “mental computational mechanisms.” We suggest that this top-down path illustrates how a desired goal can shape specific mechanisms within the cognitive machinery, in service of approaching a goal or future state. Once we decide we really want something, we have impressive capabilities that can serve those wants (Sheldon, Reference Sheldon2014).

Thus, our preferred model of science is not computational reductionism, but rather consilience (Wilson, Reference Wilson1998), in which scientists coordinate multiple levels of description using appropriate organizing constructs. We are interested in every level of analysis, from the social and interpersonal to the mechanistic. Of course, computational models may emerge as important research tools, but they do not “replace” or fully explain other levels of description. As Ryan and Deci (Reference Ryan and Deci2017) argued, psychological theories are not distinct from biological accounts, and can be coordinated with them. Yet psychological events are lawful and important and are “typically the most practical level at which we can intervene in human affairs (p. 7).” In contrast, Figure 2's mechanistic model provides little practical leverage for affecting real-world behaviors.

In conclusion, although the target article's point is well-taken (beware of over-reifying concepts), we think it is a mistake to throw out higher motivational constructs altogether. These are not just illusions or post-behavioral constructions; they reflect real causal propensities and persistent regularities in the dynamics of human striving. They help us understand both what people are trying to do in life and the social conditions that support or thwart these motives. Without them, we may be stranded in a world of mechanisms, having lost sight of the real people who deploy them.