Playing With Matches: Preparatory Cognitive Processing Shapes Affective Evaluation

Affect is linked to cognitive control

In addition to success, affect is linked to the amount of cognitive control required to complete a task (for a review, see Inzlicht et al., 2015). Difficult tasks are more likely to require more cognitive control—often because cognitive conflict must be resolved—and produce more errors, which is associated with less positive affect. Accordingly, in tasks that bear more or less on conflicting information, such as in naming font color of semantically congruent or incongruent color words (Stroop, 1935), congruent stimuli trigger relatively more positive affect than incongruent ones—even when errors are excluded (Dreisbach & Fischer, 2012). Hence, positive affect is not binarily determined by success but also modulated by the cognitive control that was associated with this success (Wirth et al., 2018; Wirth & Jall, 2026).

But then, humans sometimes do enjoy successful completion of tough, difficult and complex tasks, i.e., tasks that necessarily recruit heightened levels of cognitive control. Accordingly, and at first sight in contradiction with previous findings, correctly responding to incongruent Stroop stimuli in one study elicited more positive affect than responding to congruent Stroop stimuli, but only when participants could take their time and respond to the Stroop stimuli, thereby successfully resolving the conflict (Schouppe et al., 2015). However, in the former study, participants only briefly looked at the Stroop stimuli for 400ms and then affective priming was administered without any response to the Stroop task in the first place (Dreisbach & Fischer, 2012). Thus, while control itself is experienced as negative, conflict resolution may be experienced as positive (Inzlicht et al., 2015; Ivanchei et al., 2019), implying that conflict-loaden tasks have the potential to induce positive affect as well. However, conflict resolution needs time and resources. That might exactly be why, under certain circumstances, incongruent Stroop words elicited more positive affect than congruent ones, but only when affect was measured after enough time had passed (Fritz & Dreisbach, 2015).

Affect might be linked to preparatory processing

Unfortunately, time and resources are scarce sometimes, so that problem solvers are not always able to resolve conflicts. What can we do when conflicts cannot be resolved on the spot? Think ahead and resolve conflicts before they occur! Our mind is an avid predictor (e.g., Clark, 2013) and predicting incoming sensory input before it occurs could circumvent a negative affective response. Similarly, proactive control refers to keeping goal-relevant information in mind before “the occurrence of cognitively demanding events, to optimally bias attention, perception and action systems in a goal-driven manner” (Braver, 2012, p. 106; also see Miller & Cohen, 2001). Proactive control allows us to anticipate and prepare for effortful processing, frontloading or possibly even circumventing cognitive load.

In the present paper, we focus on one specific way to engage proactive control via biasing attention: altering the saliency of response categories. When humans make binary classifications such as whether a statement is “true” or “false”, two stimuli are “same” or “different”, or a searched for object is “present” or “absent”, typically one of these alternatives is more emphasized or salient than the other (Weeks & Proctor, 1990). Which of the two alternatives is more salient is to some extent a population stereotype, such as “right” being more salient than “left” (Olson & Laxar, 1973). However, categorical saliency can also be manipulated via instructions (Pauly & Wentura, 2025). Often the more salient alternative is the one that participants choose more quickly, which suggests adaptive preparatory processing, i.e., proactive control. For example, in visual search, response times (RTs) are typically lower when a searched target object is present in a display rather than absent, even when only a single target or a single distractor is shown (e.g., Treisman & Gormican, 1988). Similarly, participants are faster to decide that two stimuli match rather than mismatch (the so-called fast match effect, Bamber, 1969, see also Cousineau et al., 2023).

A similar way to bias attention relates to preparatory processing of certain stimuli rather than response categories. For example, an obscured object was detected faster when the contour of that object was briefly shown before the actual object appeared compared to when the contour of a different object was shown before (Reber et al., 1998). Interestingly, the objects that matched the contour were also liked more, which suggests a link between preparatory processing and affective response, possibly due to increased perceptual fluency. Similarly, participants preferred repeating stimuli over novel stimuli (mere exposure effect; Wilson, 1979) or preferred stimuli close to a prototype over more deviant stimuli (Winkielman et al., 2006). Along these lines, it was also proposed that when the activation of a schema—which we count as preparatory processing in our terminology—fits with incoming sensations, positive affect follows (Mandler, 1982). Interestingly however, a mismatch between schema and sensation was proposed to also evoke positive affect if the incoming sensations can be matched with an alternate schema. Yet another angle on preparing for the processing of certain stimuli is provided by predictive coding. When participants were asked to categorize stimuli as “face” or “no face,” neurophysiological evidence suggests that a face-specific top-down template was created against which incoming sensory evidence was compared (Summerfield et al., 2006). Analogously, when participants were asked to categorize stimuli as “houses” or “no houses”, a different but possibly house-specific top-down template was created.

In sum, humans prepare for what may come up, both over time, as in mere exposure, schema acquisition, and population stereotypes, or more spontaneously, as in priming, goal-directed proactive control, and creation of “search templates” in predictive coding. Such preparation can then lead to less effortful and more fluent processing as well as less cognitive control when sensory input matches the prepared template. Ultimately, we conjecture that solving a task with adaptive preparation should then be associated with relatively more positive affect than without preparation, as each of the aforementioned processes—increased fluency, decreased cognitive control, successfully matching template and sensory input—had previously been associated with positive affect.

Current Study

In the present study, we used a categorical saliency approach to test how preparatory processing shapes the affective experience following task completion. We conjecture that when a task-relevant stimulus corresponds to the salient pole of a category (e.g., “right” or “top”), processing of this stimulus is partially front-loaded and responding to the stimulus elicits more positive affect than when a task-relevant stimulus corresponds to the less salient pole of a category (e.g., “left” or “bottom”). We assessed (in Experiment 1) and manipulated (in Experiment 2) according to which categorization scheme participants processed physically identical stimuli and measured affect briefly after task completion. We conjectured that stimuli that matched the respectively more salient category of the currently implemented scheme are associated with more positive affect.

Specifically, we let participants use one of two categorization schemes to categorize the relationship between two digits: (1) “match” vs “no match” or (2) “sequence” vs “no sequence”. Both numbers either perceptually matched (e.g., “3” and “3”) or mismatched (e.g., “3” and “4”). Yet, if they mismatched, they always formed a sequence (e.g., a “3” is always followed by a “4”). When categorizing according to sameness, “match” is the more salient category as compared to “no match” (Bamber, 1969; Walker & Cousineau, 2019). When categorizing according to sequentiality, “sequence” is arguably more salient than “no sequence”. Consequently, the same two digits that belonged to the more salient category of one categorizations scheme (e.g., “3” and “4” when looking for “sequences”) always belonged to the less salient pole of the respectively other scheme (e.g. “3” and “4” form “no match”).

In Experiment 1, participants were free to use whichever categorization rule—sameness or sequentiality—they preferred. We assessed the employed rule afterwards and explored whether the respectively more salient response option comes with more positive affect than the respectively less salient option. Additionally, we investigated whether such affective differences also emerge when both sameness and sequentiality are clearly task-irrelevant because participants had to identify the color of matching or sequential stimuli instead. In Experiment 2, we then explicitly instructed participants to rely on either the sameness or sequentiality rule, expecting this to be a way to manipulate (rather than infer) preparatory cognitive processing, which should then have downstream consequences on affective responses. In both experiments, affective responses were measured with affective priming (Fazio, 2001; Fazio et al., 1986).

Materials and Methods

Stimuli, Procedure and Task

In total, participants engaged in 320 trials that each consisted of a prime task and a probe task. In one block, consisting of 160 trials, the prime task was the color task. In another block, consisting of 160 trials, the prime task was the digit task. Preceding each task block, participants engaged in 6 practice trials that could be repeated if deemed necessary. The probe task was a variant of affective priming (Fazio et al., 1986). Which prime task participants engaged in was separated by experimental halves. The order of prime tasks was counterbalanced between participants. We measured response time (RT) and accuracy in both tasks. The trial setup is depicted in Figure 1.

Figure 1

Prime tasks

In both prime tasks, participants first saw a digit (1–5) in gray on the left side of the screen (e.g., left digit: “3”). After 500 ms, a second digit appeared on the right side of the screen (1–6). This digit either matched the first number (right digit: “3”) or formed a sequence with the first number (right digit: “4”). This ensured first, that the predictability of specific stimuli (which digit) in match and non-match trials were identical (e.g. after a “3” another “3” or “4” were equally likely). More importantly, this allowed to categorize the digits according to their perceptual match (match or non-match) or sequential order (sequence or no sequence). Simultaneously with the onset of the second digit, both digits turned either blue or orange. In the color task, participants were to judge via keypress whether the digits turned blue or orange. We asked participants to “[…] indicate the color of the digits. Please press [Q, S] if the color is [blue, orange] and [A, D] if the color is [blue, orange]”. In the digit task, participants were to judge via keypress whether the digits form a match or a sequence. We asked participants to “indicate the identity of the digits. Please press [Q, S] if both digits [form a sequence, are identical] and [A, D] if both digits [form a sequence, are identical]!”. Half the participants responded to the color task with the keys “q” and “a” and to the digit task with the keys “s” and “d”, the other half vice versa. Which of these keys were assigned to which response (color task: blue or orange; digit task: match or sequence) was counterbalanced between participants. A “too slow” error message appeared if participants responded more than 2000 ms after onset of the second digit.

Probe task

After a subsequent response-stimulus-interval (RSI) of 300 ms, a positive or negative word was presented in the center of the screen. Participants were asked to categorize the word as quickly as possible via mouse button press as either positive or negative. The assignment of left and right mouse buttons to positive and negative was counterbalanced. As in the prime task, a “too slow” error message appeared 2000 ms after word onset. The trial concluded with an inter-trial-interval of 1200 ms.

Probe task stimuli

For the probe task, we selected 40 positive and 40 negative words from the CELEX lexical data base (Baayen et al., 1995) with a focus on maximizing extreme valences, and simultaneously matching positive and negative words regarding to arousal, frequency of appearance per million words in the CELEX corpus, and word length (Warriner et al., 2013). After the matching procedure, positive and negative words differed on the valence dimension; scaled from 1 to 9, M_positive = 7.04, M_negative = 3.01, t(78) = 31.16, p < .001. Positive and negative words did not differ with respect to arousal, frequency and word length; all |t| (78) < 1.1, all p > .30.

Questionnaires

After each experimental half, i.e. after completing all color task trials and all digit task trials, respectively, we asked participants how much they agreed to the following statements on a 5-point Likert scale from strongly disagree to strongly agree: “The task in the last block was fun.”, “The task in the last block was demanding.”, and “I was frustrated by the task in the last block.”. We also asked participants to indicate “how you are currently feeling” by picking one of nine self-assessment manikins (SAM scale; Lang, 1980). At the end of the experiment, we asked for demographic variables and about the strategy when answering sequential digits in the digit task (“What was your strategy when answering whether the digits form a sequence (for example: 5 and 6)? Did you find the answer by confirming that the digits form a sequence (when counting up from 5, it is 6)? Or did you find the answer by excluding a match (when both numbers are not identical, it is always a sequence)?”).

Results

Omnibus Anova

In line with our expectations, the ANOVA identified a Prime Task × Prime Stimulus × Probe Valence interaction; F(1, 96) = 6.70, p = .011, η²_G = .05 (see Figure 2). Full ANOVA results can be inspected in Table S1.

Figure 2

Discussion

In Experiment 1, we tested how matches of task-relevant and task-irrelevant stimulus features are evaluated on an affective level. To do so, we presented two colored digits in temporal succession. We designed these digits so that the color would always match, and we manipulated whether the digit would match or mismatch. If digits perceptually matched, they formed no sequence, whereas they did when they perceptually mismatched. Between blocks, we manipulated which stimulus dimension (color vs. digit) was task relevant.

In the color task, the task-relevant feature (color) always matched between the two stimuli, so there were no mismatches. These matches seem to have been evaluated overall as relatively positive, as positive words were overall answered faster than negative words in these blocks, though it should be noted that the positive evaluation could also be due to the positive baseline mood in Western participants (e.g., Diener & Diener, 1996). The task-irrelevant feature (digit) did not shape evaluation in a meaningful way.

In the digit task, the task-relevant feature (digit) could match or mismatch between the two stimuli. Here, we found that stimulus matches lead to a more positive evaluation than mismatches. Specifically, the RT benefit for subsequent positive probe words over subsequent negative probe words was larger after matches than after sequences. The task-irrelevant feature in this block (color) again did not modulate affective evaluation in a significant way. Overall, it seems that task-irrelevant feature matches do not whereas task-relevant feature matches do alter affective evaluation.

Further, we found that this result was modulated by the participants’ reported strategies. Specifically, participants were asked which strategy they used to complete the task, differentiating between those who check whether the stimuli are matching (match seekers) and those who check whether the stimuli form a sequence (sequence seekers). Interestingly, match seekers exhibited the effect in the expected direction. For sequence seekers, the effect was not only reduced, but descriptively even reversed, indicating that they seem to experience more positive affect after sequences instead of after matches.

These results suggest that not the stimulus match per se is experienced as positive, but the fit to the more salient pole of a categorization rule (i.e., “match” when categorizing according to perceptual similarity, and “sequence” when categorizing according to sequential order). Our overall observed positive affect following task-relevant stimulus matches was driven by the fact that most participants pursued the strategy of checking for matches rather than sequences (which is in line with earlier observations; section 3.2 in Chetverikov & Kristjánsson, 2016).

Materials and Methods

Results

Hypothesis Testing

H1 is confirmed. Match seekers experienced more positive affect after solving digit tasks with matching compared sequential digits, t(96) = 4.14, p < .001, d = .42, 95% CI_d = [.21, .61], Δm – Δs = 26 ms, Figure 3a.

Figure 3

Results regarding H2 are ambiguous. Based on our a priori analysis plan and a .05 alpha level, H2 is rejected. Sequence seekers did not experience more positive affect after solving digit tasks with sequential compared to matching digits, t(96) = –1.42, p = .079, d = –.14, 95% CI_d = [–.35, .06], Δm – Δs = –9 ms, Figure 3b. However, since the results tend to go in the hypothesized direction, we decided to follow up with exploratory analyses (see below).

Discussion

In Experiment 2, we manipulated the task set and thus the content of our participants’ proactive control by asking participants to either look for identical (“match seekers”) or sequential (“sequence seekers”) stimuli. Note that we did not manipulate the control mode or amount of proactive control. Our analyses indicate that such preparatory processing plays a crucial role in the affective response following task completion. Specifically, match seekers exhibited more positive affect after indicating that digits were identical rather than sequential. Conversely, sequence seekers exhibited more positive affect after indicating that digits were sequential rather than identical.

In Experiment 1, we showed that perceptual matches elicit more positive affect than perceptual mismatches. With Experiment 2, we now can add that it is not sufficient that task-relevant stimuli perceptually match to positively modulate affect. Instead, preparing for task-relevant stimuli seems to be the mechanism driving the effect. Results of Experiment 2 thus suggest that the main reason why identifying perceptual matches in Experiment 1 elicited more positive affect than identifying sequences was because naïve participants are more likely to expect matches than sequences.

Where is the affect coming from and why is it relevant?

So far, we have argued that preparatory processing can drive affect. How could the mechanistic details of such affect-inducing preparatory processing look like? As of yet, we sympathize with a predictive coding interpretation in which our categorical saliency manipulation led to a preparatory creation of different top-down templates against which the perceptual input was being compared against. For example, after seeing a “3”, sequence seekers would create a “4” template, which they would then compare against the actual sensory input once it appeared. Correct anticipation of the sensory input—i.e., if a “4” indeed appeared—might have then elicited positive affect (Chetverikov & Kristjánsson, 2016). Given the innate tendency of living beings to avoid negative and seek positive affect (e.g., Panksepp, 2004), this causal chain would relate to reinforcement learning (Trapp, 2017), surprise minimization behavior, and behavioral preference for template-confirming stimuli.³

If one subscribes to the notion that incorrect anticipations—i.e., when templates do not match sensory input—elicit more negative affect than correct anticipations, our predictive coding interpretation would also align with theories emphasizing on the role of affect in regulating cognitive control (e.g., Inzlicht et al., 2015). These theories state that cognitive conflict—here evoked by incorrect anticipations—elicit negative affect, which serves as a recruiting signal for cognitive control and ultimately can be used to create future templates, upregulate attention, and adjust behavior. In other words, cognitive control might not only be recruited after errors, but generally after unexpected events. Accordingly, post-error slowing, a prominent marker for cognitive control, was also found after correct answers, if correct answers were infrequent (Notebaert et al., 2009). Following a predictive coding interpretation, such a finding is not surprising, because any deviations from expectancy should recruit cognitive control, irrespective of whether the outcome is positive or negative in itself (which is also acknowledged by Notebaert et al., 2009, p. 278; also see Holroyd & Coles, 2002; Proulx et al., 2012).

The link between affect and cognitive control can also be framed in terms of effort. Cognitive control requires effort, which is frequently aversive (Kool et al., 2010). Accordingly, the less effort one needs to complete a cognitive operation, the more monetary loss one is willing to accept (effort discounting; Westbrook et al., 2013; for a review, see Kool & Botvinick, 2018). Similarly, the less fluent a stimulus is perceived, the less it is liked (Reber et al., 1998). In sum, we argue that the affective aftermath of preparatory processing can have a vast influence on how we solve tasks and how we engage with our environments, though the mechanistic details remain speculative at this point.

Does a feature need to be task-relevant to be linked to positive affect?

Perceived stimuli usually possess a large range of features like shape, color, size, location, or orientation. When compared to other stimuli, some of these can be identical and others are bound to differ. In fact, two separate stimuli can be identical regarding all features. Even if two objects were perfectly identical in every facet, they can never occupy the same space at the same time. Consequentially, problem solvers usually do not attend to all features, but only those that either are, or at most closely resemble, task relevant information. Congruent with this line of thought, we found that the digit identity feature was only relevant for affective processing when the task was related to digit identity, not when it was related to digit color.

Why does task relevance play a role? We speculate that task relevance guides attention to certain stimuli and their features, which then enables the creation of “search templates” during preparatory processing. At first glance, such an interpretation is at odds with the observation that repeated presentation of unattended stimuli can also lead to increased positive evaluation of these stimuli (mere exposure effect; Wilson, 1979). But note that the mere exposure effect can be modulated by selective attention: mere exposure to stimuli outside of selective attention will produce little effect (Yagi et al., 2009). Consequentially, one possibility is that while task relevance is not mandatory for the induction of positive affect, it is highly suited for recruiting selective attention, which might be a prerequisite for the creation of search templates during preparatory processing. Using search templates then has downstream effects on affect when compared to sensory input.

Limitations and Future Research

We took great care during the implementation of our affective measure. However, affective priming only allows us to measure affect at a single point in time. In our case, the affective measure is based on affective stimuli presented 300 ms after response. Consequentially, we are curious about the affective dynamics both after and, maybe even more interestingly, before this point estimate. After all, the idea behind preparatory processing is to front-load effort, which could mean that negative affect is not avoided but simply occurs earlier. Contrarily, it is conceivable that front-loaded effort has no affective consequences but still reduced conflict during stimulus und response processing. Clearly, a better understanding of such dynamics is desirable.

Regarding our affective measure, future research could also further clarify the measured facets of affect. Evaluative priming measures, including the presently used affective priming paradigm, are still incompletely understood (e.g., Eder et al., 2012; Rohr & Wentura, 2022), though they are used as “affective” measures (e.g., Dreisbach & Fischer, 2012). But affect is multifaceted. Current evidence suggests close ties between evaluative priming and semantic “cold” affect and looser if any ties to physiology-related “hot” affect (Rohr & Wentura, 2022; others prefer the terms semantic and affective valence, respectively, Itkes et al., 2017; or core affect for hot and “perception of affective quality” for cold affect, Russell, 2003, p. 157). Thus, it is possible that while solving the prime task, participants categorized some parts of the solving process as positively or negatively valent, which then leads to downstream congruency effects regarding the probe word categorization, without involvement of “hot” affect. That being said, we deem our results relevant irrespective of whether affect materialized in the “hot” variant, the “cold” variant related to activation of affective knowledge, or both. Both variants would indicate relevance for information processing, decision making, and ultimately shape behavior (e.g., Russell, 2003, pp. 157–158).

Lastly, we cannot disambiguate whether template-confirming stimuli generated positive affect, or template-deviant stimuli generated negative affect, or both. A disambiguation in future studies would be desirable because it might bring us closer to the causes of the affective response. On the one hand, a reduction in negative affect could be caused by a reduction of effort, less cognitive control, or less conflict evoked by mismatches between templates and sensory input. On the other hand, positive affect could serve as reward signal for matches between templates and sensory input.

Finally, future research should pinpoint the event that excites the affective signal. In our setup, it could be that the presentation of an expected stimulus (i.e., a 3 following a 3 for match seekers) triggers the relatively more positive signal. Alternatively, it might be the decision to categorize the stimuli as a match, or it might be the overt response of stimulus categorization. A more detailed understanding of when the affective signal is elicited will also highlight the point in time in which anticipatory processes are evaluated.