Why do self-ratings usually rise when a task gets harder?
Cognitive load is not directly visible, so researchers infer it indirectly through several channels: what people report, how they behave, how they perform, and physiological signals [2]. Subjective ratings are the simplest channel: after a task, a person says how much mental effort it took. When researchers deliberately make a task harder or easier, those ratings tend to shift accordingly. In the sketching study with 7- to 12-year-olds, children reported higher intrinsic and extraneous load when the task versions were designed to impose more of each, and their ratings were substantially related to both the induced load and their performance [1].
That pattern is what makes ratings attractive in instructional research: they are cheap, quick, and sensitive to the difficulty a designer intentionally builds in. The same study found that the physical traces of drawing, the speed and pressure of the smartpen, also tracked induced load and performance [1]. Two different windows onto the same underlying demand, in other words, both showed the expected response.
If both track difficulty, why don't they agree with each other?
Here is the surprising part. In the sketching study, the children's ratings and their smartpen measures were each related to load and performance, but the two were not substantially related to one another [1]. That means a child who reported high load was not necessarily the child whose pen moved slowly or pressed hard. The authors concluded that both approaches hold potential for measuring load and performance, but it is questionable whether they measure the exact same constructs [1].
This is a useful distinction for anyone designing an experiment. A rating is a person's summary judgment about their own experience; a pen-speed or pressure measure is a behavioral byproduct recorded while they work. They can both respond to the same manipulation without being interchangeable. Treating one as a stand-in for the other would blur what each is actually telling you.
What do objective measures add that ratings miss?
Objective measures are not a single family. They include performance outcomes, behavioral traces, physiological signals, and secondary-task responses, and researchers have proposed ways to compare them on a common scale, such as a signal-to-noise ratio, so that different measures can be judged by how well they detect changes in load [2]. In an n-back task study with 28 participants, this comparison framework was applied across measures from four broad groups, and the authors reported that the method can be useful for assessing individual measures of cognitive load [2].
Other work has argued that objective measures may have an advantage over subjective ones in some respects [4], and dual-task paradigms, which use reaction times and errors on a secondary task as a load signal, come with their own limitations for capturing cognitive load [5]. The practical upshot is not that one family wins. It is that each measure carries different information, and the choice depends on what question you are asking.
How should instructional researchers use ratings alongside performance?
The most defensible reading of these studies is to use subjective ratings and objective measures as complementary evidence rather than as checks on each other. Ratings tell you how demanding the task felt and respond to designed differences in load [1]. Performance tells you how well the task was done, and behavioral or physiological measures can capture load-related changes that a person may not consciously report [2]. When they diverge, that divergence is itself informative: it may mean the measures are tapping different constructs, as the sketching study suggests [1].
Scope matters here. The clearest direct comparison of ratings and an objective behavioral measure comes from a small laboratory study of 36 children doing sketching tasks [1], so the finding that the two were not substantially related is best read as evidence about that age group and task type rather than a universal rule. The broader comparisons of measure types come from adult n-back tasks [2] and from conceptual discussions of what subjective measures can reliably capture [4][5]. Across these contexts, the consistent message is that subjective ratings agree with performance in direction but should not be assumed to agree with objective measures in detail.
Sources used in this answer
Digital ink and differentiated subjective ratings for cognitive load measurement in middle childhood
In 36 children aged 7–12 doing standardized sketching tasks, subjective cognitive load ratings and smartpen velocity and pressure measures were each substantially related to induced load and performance, but the ratings and smartpen measures were not substantially related to each other.
Comparison of concurrent cognitive load measures during n-back tasks
A signal-to-noise-ratio performance metric was proposed and used to compare cognitive load measures from four broad groups on the same scale using data from 28 participants performing n-back tasks of varying difficulty, and the method was reported as useful for individually assessing different measures.
T-MOOC, cognitive load and performance: analysis of an experience
The study examines cognitive load in a t-MOOC instructional setting using a validated scale and relates it to performance outcomes, noting that both objective and subjective methods exist for measuring load.
Subjective measures of cognitive load: what can they reliably measure?
The paper examines the reliability of subjective cognitive load measures and compares them with objective measures, suggesting objective measures may have an advantage over subjective measures in some respects.
A current view on dual-task paradigms and their limitations to capture cognitive load
The paper discusses dual-task paradigms as objective cognitive load measures based on reaction times and interval errors, and outlines their limitations for capturing cognitive load in instructional settings, contrasting them with subjective ratings.