Visual perception of lifted weight

Is it possible to visually perceive the weight of an object? Intuitively, this does not seem possible. After all, what an object looks like tells you nothing reliably about what it is made of. It is only by attempting to carry or lift an object can one determine its weight or mass. It would hence seem that knowledge of object weight is one that is reserved only for the haptic, and not the visual, system.

The research

Runeson & Frykholm (1981) argue otherwise. Object weight CAN be visually perceived! In this paper, they define the study of motion in two ways. Firstly, motion and its changes can be accounted for descriptively. This is done by studying kinematic variables, defined as those of position, time, and the temporal derivatives of position (e.g., velocity, acceleration). Given that any real-world task of interest to psychologists typically involves human behaviour (and this entails some sort of motion), studying kinematics is a good starting point to describe the motions and changes in motion involved. However, kinematics are purely descriptive -- they tell you nothing about what caused those motions and their changes. To account for this, we need to study the dynamics of a task, which deals with kinematic variables plus mass, hence allowing scientists access to the different forces involved in the subsequent kinematic patterns observed.

We now face a problem. If we define 'dynamics' more broadly to mean any variable that has an influence on how a real-world event pans out, then as biological perception-action systems, we want to perceive the dynamic properties of the world. However, our visual perceptual system only has access to kinematics (think about watching someone lift a box -- you only have access to the resulting kinematic motions, and yet you somehow have an inkling of the weight of the box and the forces required to lift that box). Can we perceive the dynamic influences of events just by detecting kinematic variables?

Can people visually perceive the weight of a lifted box?

Runeson & Frykholm (1981) tackled this question first by conducting a series of pilot studies where they got a bunch of actors to lift boxes that varied in weight. They then got another group of observers to watch those actors and to report what they thought the weights in the boxes were. The overall results showed that observers were able to accurately discriminate the box weights, with some participants even showing high levels of precision.

So, we've established that yes, the phenomenological (i.e., first-hand) experience of knowing the weight of a box (i.e., the dynamics) simply by watching someone lift it can be replicated under experimental conditions. However, the interesting part is what came next. When asked what information they used to determine the weights, participants mostly cited facial expressions, observable muscular strains, and sounds from the actor or box. Crucially, only some of them cited kinematic properties (i.e., the position, velocity, and acceleration of different parts of the actor's body throughout the lifting process).

Does this mean that kinematics aren't actually useful as information about dynamics? Not quite. Enter experiment 1, where the authors aimed to see if exposure to only kinematic information would replicate the results from the pilot studies.

Visual kinematic patterns specify dynamic weight

The methodological challenge of experiment 1 was to remove information like facial expressions, muscular strains, and auditory cues, while still preserving the kinematics of the actor. To do this, Runeson and Frykholm used a point-light display paradigm. Actors wore dark clothing and bright patches of tape stuck to their joints. These retroreflective tapes were also attached to the corners of the lifted box (see Fig. 1). They were then filmed carrying boxes of different weights. Consequently, observers viewed these video recordings and made judgements of the box weights.

Fig. 1 Locations of retroreflective tapes (Runeson & Frykholm, 1981)

The results (see Fig. 2) showed that, despite only having access to kinematic information, participants performed really well in this task, with their judge weights increasing linearly with actual weights, with an average beta coefficient of 0.87. To briefly explain this statistic, it just means that with every 1 kg increase in actual weight, participants' judged weight increased by around 0.87 kg. Perfect performance would (roughly) equate to a value of 1.00.

Fig. 2 Experiment 1 results (Runeson & Frykholm, 1981)

A second experiment was conducted to see how performance would change when (a) full visual and auditory information was restored and (b) haptic information was also present. This time, participants observed another person lift boxes in real time and under normal visual conditions. On separate trials, the same participants also got to lift the boxes themselves. Results (see Fig. 3) showed average slopes of 1.00 and 1.20 under the visual and haptic conditions, respectively. That is, compared to experiment 1, participants performed better in the visual condition and tended to overestimate the weights in the haptic condition. This suggests, firstly, that visual perception of lifted weights is comparable to that with haptic perception. Secondly, while performance is best under full visual conditions, kinematic variables still provide the most useful information for the perception of dynamics (as compared to facial expressions, auditory cues, and muscular strains). 

Fig. 3 Experiment 2 results (Runeson & Frykholm, 1981)

There are some results not mentioned here. For example, dispersion was higher, and participants tended to understimate the weight range when only kinematic patterns were made available (compared to the full visual condition). However, it's important to note that while absolute scaling (i.e., judgement of absolute weight values) suffered slightly, performance of relative scaling (i.e., heavier weights judged as heavier than lighter ones, regardless of the absolute value) remained high. This lends support that kinematic patterns do carry information specifying dynamic properties (i.e., box weight) of the event (i.e., person lifting the box).

That being said, all this study has done is to establish that there are sources of kinematic information that one can use to perceive box weight. What the exact informational pattern is, we don't know (yet). These are likely higher-order kinematic patterns that remain invariant for a given box weight. Box weight also only accounts for one part of the dynamics. Here, the authors argue that other factors, like initial lifting position, will have a hand in shaping the kinematic properties available for detection. 

References

Runeson, S., & Frykholm, G. (1981). Visual perception of lifted weight. Journal of Experimental Psychology Human Perception & Performance, 7(4), 733–740. https://doi.org/10.1037/0096-1523.7.4.733