Impression Trajectories

1946 — Impressions as organized wholes

In his foundational work, Solomon E. Asch showed that swapping "warm" for "cold" in an otherwise identical list of traits reshaped an entire impression of the person those traits described. Those who heard "warm" (“cold”) were far more likely to also picture the person as generous (ungenerous), despite generosity never being mentioned. Changing the order of the trait list shifted impressions too. These findings established that forming impressions is a holistic, organizing process, where a trait's meaning depends on everything around it.

In Asch’s experiments, participants described the person in their own words, preserving much of the richness of real impressions. But at the time, interpreting how impressions differed came down to his own judgment, not objective measurement. And for all that richness, the person judged was nothing more than a list of traits on a page, not someone actually observed doing or saying anything. Fittingly, the paper's own closing pages looked to the future, when we might instead study real people "in movement and growth."

1960s–80s — Integration and memory

In the decades that followed, researchers turned Asch's holistic insight into something you could calculate. Norman Anderson developed what became known as "cognitive algebra," principles by which social information combines through weighted averaging, with earlier information typically counting for more. Around the same time, other researchers found that people told to form an impression of someone from a list of their behaviors later recalled those behaviors more accurately than people told simply to memorize the same list. This was taken as evidence that forming an impression means actively linking newly observed behavior to prior ones already encountered, and that it's this organized structure, not the separate facts themselves, that later judgments about the person actually draw on.

Still others distinguished impressions formed “online,” as information arrives, from those reconstructed later purely from memory, showing people could reach a stable evaluation of someone well before ever being asked to report it. For all this precision, though, these advances relied on data collected once an evaluation was already complete, or memory probes well after the encounter, meaning the integration process was inferred backward.

1990s — Context shapes meaning

By the 1990s, a new generation of computational models set out to explain not just how traits combine, but why context seems to change their meaning. These connectionist models proposed a single mechanism behind many earlier findings: perceivers were argued to settle on an interpretation the way a network settles into a stable state, with every piece of information (e.g., traits, behaviors, prior beliefs) simultaneously pulling on and constraining one another until a coherent picture emerges.

Under this view, no new piece of information is ever read in isolation; it's always interpreted in light of everything a perceiver already believes about the person. But it remained a largely theoretical achievement: these models were fit to existing patterns of human ratings after the fact, simulating how impressions might settle into place, rather than directly and continuously measuring that settling process as it happened in real time.

2000s–10s — Real-time signals, broad dimensions

The 2000s brought a genuinely new kind of evidence: real-time behavioral traces of impressions forming. By tracking the physical path of a computer mouse as someone made a snap judgment, researchers could see competing interpretations of facial judgments physically tugging at a person's hand before one interpretation ultimately won out. Freeman and Ambady's (2011) dynamic interactive theory built on the constraint-satisfaction models of the 1990s but extended that logic down to social perception itself, modeling how raw sensory cues activate categories and stereotypes in the first place, rather than starting from already-categorized traits. It also came with something those earlier models lacked: direct evidence that this settling process unfolds continuously, not just a model fit afterward to existing ratings. Still, this real-time signature was captured only within a single, split-second decision, not across anything resembling a real social encounter.

Around the same period, other research consolidated years of findings into a small number of broad dimensions, warmth and competence chief among them, that seemed to organize much of social judgment. It was a genuinely useful simplification, letting researchers compare impressions across many different studies, even as it left open what might fall outside a handful of predetermined categories.

Late 2010s–early 20s — Continuous tracking, one dimension at a time

In the years since, researchers found a way to bring that same real-time resolution closer to an actual social encounter. Rather than capturing a single split-second decision, they had people continuously rate one predefined trait — how trustworthy someone seemed, say — moment to moment while watching a naturalistic video unfold, showing that impressions shift meaningfully as new verbal, visual, and vocal information arrives in real time. It was the first time continuous measurement met naturalistic, unfolding behavior, rather than a single frozen instant or a split-second choice. But it came at a real cost: to track something continuously, researchers still had to fix in advance which single trait participants would rate, leaving open whether that same continuous approach could ever capture the fuller, higher-dimensional richness of how impressions actually form.

Present — Trajectories through meaning

My own work picks up this thread, extending it from fixed, researcher-provided trait lists to the first minutes of a real, unfolding social encounter — the speech, gestures, and actions people naturally read each other by. Rather than treating impressions as something combined once, from a complete set of inputs, this approach follows them as they're spontaneously and continuously constructed, capturing each new inference in the moment it forms rather than only after an encounter ends. It's an attempt to answer the same question the field has circled for eighty years — how the pieces come together — using tools that finally allow that process to be measured directly and continuously, rather than inferred backward from ratings, memory tests, or a single split-second decision. In a sense, it's an attempt at the richer field Asch himself pointed toward: impressions of real people, observed not as a finished judgment, but in movement.