Stimulus–response congruency effects depend on quality of perceptual evidence: A diffusion model account

Tomkins, Blaine

doi:10.3758/s13414-022-02642-9

Stimulus–response congruency effects depend on quality of perceptual evidence: A diffusion model account

Published: 01 February 2023

Volume 85, pages 1335–1354, (2023)
Cite this article

Download PDF

Attention, Perception, & Psychophysics Aims and scope Submit manuscript

Stimulus–response congruency effects depend on quality of perceptual evidence: A diffusion model account

Download PDF

Blaine Tomkins ORCID: orcid.org/0000-0003-4172-8711¹

469 Accesses
Explore all metrics

Abstract

Individuals often need to make quick decisions based on incomplete or “noisy” information. This requires the coordination of attentional, perceptual, cognitive, and behavioral mechanisms. This poses a challenge for isolating the unique effects of each subprocess from behavioral data, which reflect the summation of all subprocesses combined. Sequential sampling models offer a more detailed examination of behavioral data, enabling us to separate decisional and non-decisional processes at play in a task. Participants were required to identify briefly presented shapes while perceptual (duration, size, location) and response features (location-congruent/-incongruent/-neutral) of the task were manipulated. The diffusion model (Ratcliff, 1978) was used to dissociate decisional and executive processes in the task. In Experiment 1, stimuli were presented for either 20 or 80 ms to the left or right of a central fixation while response keys were positioned horizontally. In Experiment 2, stimulus size was manipulated rather than duration. In Experiment 3, response keys were positioned vertically. Results showed a duration x response mapping interaction. Participants displayed stimulus–response (S–R) congruency biases only on short-duration trials. This effect was observed for both horizontal and vertical response key mappings. Stimulus size affected participant response speed, but did not elicit S–R congruency biases. The present findings show that when perceptual quality of evidence is poor, individuals rely more heavily on spatial-motor mechanisms when making speeded choice decisions. Furthermore, positioning response keys vertically is insufficient to eliminate S–R congruency effects. Diffusion model parameters are presented and implications of the model are discussed.

A diffusion model for the congruency sequence effect

Article 08 June 2022

Response biases in simple decision making: Faster decision making, faster response execution, or both?

Article 31 July 2017

Some task demands induce collapsing bounds: Evidence from a behavioral analysis

Article 29 May 2018

People are often faced with situations requiring quick decisions or judgments based on little or incomplete information. Such situations may include any time an individual must detect, recognize, identify, or classify a stimulus in order to initiate an appropriate behavioral response. Even relatively quick decisions require an individual to draw on a combination of attentional, perceptual, decisional, and motor processes. A large body of research has shown that these processes, while distinct, can interact to influence behavior. For instance, individual performance on speeded decision tasks are generally faster and more accurate when spatial/perceptual features and response features of a task are related. Such effects, known as stimulus–response (S–R) congruency effects, have been extensively studied in cognitive psychology (Pashler, Johnston, & Ruthruff, 2001; Proctor & Vu, 2006). However, a challenge facing theoretical accounts of S–R congruency effects is the difficulty of separating effects due to perceptual and motor mechanisms, respectively, in behavioral data. Sequential sampling models can address the separable, additive, and interactive effects of perceptual, decisional, and motor processes. To that end, the goals of the current study are: (1) to examine how the quality of evidence extracted from visual stimuli influences S–R congruency effects; and (2) to utilize a sequential sampling model of perceptual decision making to assess the separate and interactive effects of perceptual/decisional and motor processes in the S–R congruency effect.

It is well known that the speed and accuracy of visual object recognition can be influenced by various features of an image, including but not limited to: size (Craddock & Lawson, 2009; Graf, 2006; Te Pas, Kappers, & Koenderink, 1996; Tomkins, 2021), location (Golomb, Kupitz, & Thiemann, 2014; Graf, 2006), luminance (Tomkins, 2021), clarity (Pegna, Khateb, Michel, & Landis, 2004; Shahangian & Oruc, 2014; Wyatte, Curran, & O’Reilly, 2012)and presentation duration (Potter, Wyble, Hagmann, & McCourt, 2014; Thorpe, Fize, & Marlot, 1996). Objects are generally more difficult to recognize when either viewing conditions or the quality of the image itself are degraded. While this phenomenon is well-established, the relationship among perceptual quality, decision processes, and motor (i.e., executive) processes is less understood. The relationship among perceptual, decisional, and motor processes is no subtle matter. The vast majority of studies examining perceptual and cognitive processes require some form of motor response from the participant. This may include a keypress, verbal response, finger point, or similar behavior. Exceptions may include EEG or eye-tracking techniques. This situation presents a challenge for interpreting behavioral data, since data reflect the cumulative effects of attentional, perceptual, decisional, and motor processes. However, there is evidence that these distinct processes do not always operate in a serial fashion, but often in parallel. That is, (pre)motor processes may initiate before perceptual processes have completed (Proctor & Vu, 2006). This is problematic since it is generally assumed that statistical differences across experimental conditions genuinely reflect the intended process(es) in question given that all other testing procedures are kept the same. However, observed differences in mean RT across different experimental conditions intended to measure some aspect of cognition may not actually reflect cognitive processes at all. There is a large body of evidence that response modality and response mapping can have significant influences on data intended to examine attention, perception, or cognitive mechanisms (Gomez, Ratcliff, & Childers, 2015; Hazeltine & Ruthruff, 2006; Kramer, Cox, Yu, Kravitz, & Mitroff, 2021; Philipp & Koch, 2005; Phillips & Ward, 2002; Roggeveen, Prime, & Ward, 2005; Taylor & Klein, 2000). For instance, spatial mapping of attentional or perceptual codes and motor processes can combine into a single response code over repeated trials (i.e., S–R congruency effect). Such findings have prompted some authors to suggest the keypress method should be discontinued in favor of alternative methods (Kramer et al., 2021). While this proposal may be extreme, there are valid criticisms of the keypress modality that require careful consideration in response time research. Furthermore, the common practice of averaging response time data has the potential to introduce artifacts (Estes, 1956; Gómez, Breithaupt, Perea, & Rouder, 2021). To address these issues, the present study systematically manipulated perceptual and response features in a speeded choice decision task. Additionally, the present study implements a mathematical model that considers the full response time distribution and assesses participant response time and accuracy in a single analysis to examine the select effects of perceptual, decisional, and motor processes in a simple decision task.

Models of stimulus–response congruency

Several theoretical models have been proposed to explain S–R congruency effects. Although they differ in specific features, current models show consensus that S–R congruency is composed of two separate effects: (1) general effects, and (2) task-specific effects. There are fundamental patterns that are shown irrespective of specific stimuli or task demands. For instance, an image presented to the right visual field will automatically attract spatial attention resources and prime a response to the right regardless of what the image is or whether it is relevant to the task demands. In contrast, there are congruency effects that vary based on stimulus features and/or the specific demands of the task. The dual-process model (De Jong, Lang, & Lauber, 1994) explicitly distinguishes between these two kinds of processes as unconditionally automatic and conditionally automatic effects, respectively, and argues the two processes operate independently. An alternative account is the dimensional overlap processing model (Kornblum & Lee, 1995), which conceptualizes stimulus and response information as vectors of features. In this view, S–R congruency effects are the result of an overlap or match between the respective attributes (i.e., dimensions) between the stimulus vector and response. The more dimensions that match between the two processes, the larger the S–R congruency effect. It is important to note that experimental tasks used to examine S–R congruency effects do not involve only perceptual and motor processes, but also decision processes. While current models have provided a great deal of insight into the dynamic relationship between perceptual and motor mechanisms, they offer little explanation concerning how the decision process unfolds. It is one thing to argue, for instance, that an individual must make a decision concerning whether a feature of a stimulus is task-relevant, but it is another to show how this decision is made and make specific predictions for data based on modeling assumptions. Purely verbal accounts are limited in this respect. In contrast, sequential sampling models can help reveal the decision-making process and make predictions for data based on modeling parameters which reflect specific psychological processes.

Diffusion model

Sequential sampling models have proven highly successful at describing and predicting human behavior in perceptual decision tasks. These are mathematical models that characterize decision making as a process in which noisy evidence from a stimulus accumulates over time until the threshold for a particular decision is met and a response is initiated. In particular, the diffusion model (Ratcliff, 1978; Ratcliff, Voskuilen, & McKoon, 2018) has reliably shown to account for participant behavior in “one shot” decisions across a variety of different contexts (see Ratcliff, Smith, Brown, and McKoon (2016) for a review). A strength of the diffusion model is the ability of the model to separate decision processes and non-decisional processes (e.g., memory processes, executive processes). Figure 1 shows a graphical representation of a hypothetical trial within the diffusion model. The response time in a speeded dual-choice task can be separated into three components: the time required to extract the physical and psychological features of a stimulus relevant to the task (encoding time), the time required for the diffusion process to reach one of two decision boundaries (accumulation of evidence), and the time required for the motor response. The sum of the encoding time and the time required for the motor response are represented by the parameter T_er, and the s_t parameter (its range).

Drift rate

In the diffusion model, noisy evidence accumulates towards a decision boundary at a variable rate (see Fig. 1). This process is referred to as drift. The average rate at which the evidence drifts towards a decision boundary is referred to as “drift rate”. Conceptually, the drift rate can be thought of as the quality of evidence extracted from a stimulus. For instance, suppose a participant must quickly decide whether a briefly presented object is a circle or a square. Further suppose that on a particular trial the shape presented is a square. If a relatively large and clear image is displayed there should be a large positive drift rate—reflecting that the object is relatively easy to recognize. However, if the image were more difficult to perceive, either by being made very small, fuzzy, or being presented for only a few milliseconds the value of the drift rate should be smaller—reflecting a more difficult decision due to lower quality evidence from the stimulus. In this case, the participant will likely take longer to make a decision and the drift is more likely to cross the “circle” threshold (represented by 0 in Fig. 1), resulting in an error.

When a participant is required to make a series of binary decisions over many trials, there are two sources of variability: within-trial variability and across-trial variability. The diffusion model accounts for both sources of variability. Within-trial variability is represented in Fig. 1 by the jagged line. As mentioned above, the accumulation of evidence within a trial is noisy (i.e., that is, variable). This parameter is scalar, which means that the same predictions could be generated by changing this parameter while scaling the other model parameters. Across-trial variability is contained within the η parameter. In most tasks, stimuli are grouped based on shared features or experimental conditions and it cannot be assumed that all stimuli have equal discriminability.

Not only can the drift rate be affected by features of stimuli, but also by bias. Participants show faster response times and better accuracy when the location of a target stimulus is location-congruent with the response key (S–R congruency effect). For instance, if the response key for “square” is positioned to the participant’s right, they will be faster and more accurate when a square is presented on the right side of a display relative to when the square is presented to the left. The diffusion model can capture this bias by considering the boundary position. The two parameters of the model that describe the boundary position are z, the location of the starting point (with a range of s_z), and a, the distance between the boundaries (with the location of the negative boundary assumed to be set at 0). When no bias is present, z = 0.5. When biases are present, z moves closer to the boundary the participant is biased towards, deviating either greater (or less than) 0.5.

The present task

The present study includes a simple shape recognition task. On each trial, participants were presented with either a square or a circle to either the left or right of a central fixation very briefly and instructed to press one key to classify the stimulus as a square and a second key to classify the stimulus as a circle. To test the selective and interactive influence of perceptual and motor processes both a perceptual and motor manipulation were included. The perceptual quality of the visual code was manipulated by varying presentation duration (Experiment 1) and stimulus size (Experiment 2). The motor response was manipulated by varying stimulus–response mapping. Specifically, the key to identify a stimulus as a “square” was positioned to the right of the participant, whereas the key to identify a stimulus as a “circle” was positioned to the left. In effect, half the trials are location-congruent (i.e., square presented to the right, circle presented to the left) and half the trials are location-incongruent (i.e., square presented to the left, circle presented to the right). In Experiment 3, response keys were positioned centrally and vertically to the display. This design enables us to tease apart the perceptual, cognitive, and motor processes involved in participant behavior and also systematically examine the selective, joint, and interactive effects among these factors.

Previous research has shown that S–R congruency effects may be decreased when stimuli are positioned vertically, rather than horizontally, on the display (Wiegand & Wascher, 2007). However, the extent to which S–R congruency effects are influenced by keeping the position of stimuli fixed and manipulating the relative position of the response keys has received less attention. This question is important since many experiments require lateral presentation of stimuli, thus positioning stimuli centrally and vertically is not an option. Furthermore, the present study goes beyond assessment of S–R effects by including a perceptual manipulation. This design provides a broader examination of choice behavior, enabling us to examine the effects of spatial-perceptual-motor processes. If S–R congruency effects are attenuated or eliminated by this vertical mapping assignment, we will have additional evidence that decisional biases are primarily motor-based. However, if S–R congruency effects are reduced depending on the perceptual quality of the stimulus, this suggests that S–R congruency effects are the result of a combination of early perceptual processes (i.e., since the manipulation is stimulus presentation duration), spatial attention, and motor mechanisms.

Hypotheses

With the design established, four competing accounts are proposed with each making specific predictions for the data:

Null account: No effect of perceptual evidence or stimulus–response mapping
Perceptual account: Decisions based solely on perceptual evidence; No effect of stimulus–response mapping
Stimulus–response account: Decisions based solely on stimulus–response mapping; No effect of perceptual evidence
Interaction account: Perceptual and stimulus–response processes interact to influence decisions

The first hypothesis can be considered the null model, where no significant differences are expected based on perceptual quality or response mapping. The second hypothesis assumes that the present task is primarily a perceptual task. Specifically, participants will show worse performance (slower RT, lower accuracy) when the perceptual features of the stimuli are degraded (i.e., shorter presentation, smaller size); however, the effects of response mapping will have no influence on performance. The third hypothesis emphasizes the motor aspect of the task. In this account, participants are expected to show S–R congruency effects, but no differences based on the perceptual quality of the stimuli. This outcome would suggest that S–R congruency effects are the result of stimulus–response mapping only. In other words, S–R congruency effects are the result of spatial attention and motor processes, while perceptual processes are not involved. The fourth hypothesis assumes that perceptual and response-mapping manipulations will interact to influence participant performance. Figure 2 shows the modeling predictions for each of the four hypotheses (see Angele, Baciero, Gómez, and Perea (2022) and Gomez and Perea (2020) for examples of this approach). There is a principled way for the model to account for the data: presentation time (Experiments 1&3) or stimulus size (Experiment 2) should only affect the drift rate parameter, while the response thresholds and the starting point of the diffusion process should reflect the response biases.

There are two potential patterns that may emerge in the interaction account, each with different implications. First, participants will display a larger S–R congruency effect when the perceptual quality of the stimulus is degraded. When the perceptual code affords insufficient evidence for a decision either due to the stimulus being presented too briefly (Experiment 1) or the stimulus being smaller (Experiment 2), participants will rely more on spatial-response biases and less on perceptual evidence. This pattern would indicate that better perceptual information helps reduce S–R effects. When quality evidence is afforded by perceptual codes, this quality evidence aids decisional and motor processes. The second pattern that may emerge from the interaction account is that participants will display larger S–R congruency effects with better perceptual evidence (i.e., longer presentation time, larger stimulus size). This pattern would suggest that perceptual information enhances S–R effects. Specifically, high-quality perceptual information interferes with decisional and motor processes, resulting in greater responding bias.

Experiment 1

Methods

Participants

Thirty subjects from the College of Saint Benedict and Saint John’s University participated in Experiment 1. Prescreening was conducted to ensure that all participants had normal or corrected-to-normal vision, and no history of visual impairments. Participants received course credit for participation.

Materials and procedure

Stimuli consisted of an image of a square and the image of a circle. Each image constituted a 1.5 cm x 1.5 cm region of the screen. Participants were positioned 48 cm from the center of the display. Each trial began with a centrally-presented fixation “+” that remained on the screen for 1000 ms (see Fig. 3). Following the fixation, a stimulus (square or circle) was briefly presented either to the left or the right of the fixation with the inner edge of the stimulus 2 cm from the center of the screen, creating a 2.38^∘ visual angle. The stimulus subtended \(\sim 1.8^{\circ }\) vertically and horizontally. On half the trials, the stimulus appeared for 80 ms (high quality). On the other half of trials, the stimulus appeared for 20 ms (low quality). As shown in Fig. 3, a backwards mask was presented immediately following the target to prevent additional evidence accumulation from sensory memory processes. Participants were instructed to indicate by keypress whether the flashed image is a circle or square as quickly as possible. When the stimulus is a square, the participant was instructed to press the “M” key (positioned to the right) on the keyboard with the right index finger. When the stimulus was a circle, the participant was instructed to press the “Z” key (positioned to the left) with the left index finger. Each participant completed a total of 480 trials broken down into six blocks of 80 trials each. Trials were fully counterbalanced such that an equal count (n = 60) of each of the 8 experimental conditions were presented to each participant: 2 stimulus (circle, square) x 2 presentation location (left, right) x 2 duration (long, short). All testing was conducted using E-Prime 2.0 testing software.

Results

Behavioral results

Prior to analysis, raw RTs shorter than 100 ms or longer than 2000 ms were removed (1.7 % of trials). With outliers removed, separate linear mixed-effects regressions were conducted for correct RTs and error RTs including stimulus duration (short, long) and response mapping (location-congruent, location-incongruent) as fixed factors. To avoid the issue of nonlinear transformations for response time data, a gamma distribution was specified for all response time analyses. Accuracy was analyzed with a logistic mixed-effects regression including perceptual quality and response mapping as fixed factors.

Accuracy

Results (see Fig. 5) showed a significant main effect of duration, OR = 0.07, CI = 0.06, 0.09, p < .001. Participants were more likely to make an error on short-duration trials (P = 0.33) than long-duration trials (P = 0.04). A significant main effect of response mapping was also found, OR = 0.67, CI = 0.52, 0.85, p < .001. The probability of an error was larger on incongruent trials (P = 0.20) than congruent trials (P = 0.16).

Correct RT

For correct trials, a significant main effect of duration was found, b = -164.55, p < .001 (see Fig. 4). Response times were shorter on long-duration trials (M = 436 ms) than short-duration trials (M = 624 ms). The main effect of response mapping was also significant, b = -7.31, p < .001. Participants responded faster when targets were location-congruent with the response key (M = 521 ms) than location-incongruent (M = 538 ms) Lastly, a significant duration x mapping interaction was observed, b = 17.53, p < .001. To further examine this interaction effect, a post hoc (Tukey) test was conducted (see Table 1). On incongruent trials, response times were faster in the long-duration condition than the short-duration condition, b = -182.09, p < .001. On congruent trials, response times were also faster in the long-duration condition than the short-duration condition, b = -164.55, p < .001. On short-duration trials, response times were shorter in the congruent mapping condition than the incongruent condition, b = -24.85, p < .001. However, no significant difference based on mapping was found on long-duration trials, b = -7.31, p = .07.

Table 1 Mean response times (correct trials)

Full size table

Error RT

On error trials, a significant main effect of duration was found, b = -91.23, p < .001 (see Fig. 4). Incorrect responses were made faster in the long-duration condition (M = 518 ms) than the short-duration condition (M = 659 ms). There was also a significant main effect for mapping, b = -64.52, p = .01. Errors were made faster on incongruent trials (M = 620 ms) than congruent trials (M = 672 ms). The duration x congruency interaction effect was also significant for error trials, b = -33.13, p < .001. Post hoc tests showed that on congruent trials, errors were made faster in the long-duration condition than short-duration condition, b = -91.2, p < .001. Second, on incongruent trials errors were also made faster in the long-duration condition than the short-duration condition, b = -124.73, p < .001. Third, in the long-duration condition error RTs were faster when the stimulus was location-incongruent than when the stimulus was location-congruent, b = -64.5, p < .001. Lastly, in the short-duration condition error RTs were faster in the incongruent condition than the congruent condition, b = -31.1, p = .007.

Model fit

The model was fit to the data using the D*M method (Van den Bergh, Tuerlinckx, & Verdonck, 2020). The model was used to fit correct and error RTs across the eight experimental conditions: two stimuli conditions (square, circle); two durations (short, long); and two response mapping conditions (location-congruent, location-incongruent). All parameters were allowed to change freely across the different duration x mapping conditions; however, within each duration x mapping condition only the drift rate was allowed to vary. Table 2 shows the parameter estimates derived from the model.

Table 2 Diffusion model parameters

Full size table

Discussion

In Experiment 1, participants completed a speeded decision task requiring them to identify a stimulus as a square or circle. Stimuli were presented either on the same side as the response key or on the opposite side. Additionally, stimuli were presented either very briefly (short condition) or for a longer duration (long condition). Participant data were then entered into a diffusion analysis and parameter estimates were derived. The discussion will be separated into two parts. First, the behavioral results will be discussed, followed by examination of the model parameters.

Participants made more errors when stimuli were presented for only 20 ms compared to 80 ms (Fig. 5). Consistent with the perceptual hypothesis, this finding confirms shape recognition was more difficult with shorter exposure duration. Participants also made more errors when the stimulus appeared on the opposite side of the display (S–R incongruent) from the response key. This result supports the stimulus–response hypothesis. Interestingly, the data did not reveal a duration x mapping interaction for accuracy. Thus, the accuracy results suggest that both perceptual and motor processes are involved in the present task, but these respective processes may operate independently.

Unlike the accuracy results, the RT data did reveal a duration x mapping interaction effect. As shown in Table 1, participant RTs on correct trials showed little difference based on mapping position when stimuli were presented for longer. On short-duration trials, however, mean RT was significantly shorter when the stimulus was location-congruent with the response key. This pattern is consistent with the interaction hypothesis, suggesting participants relied more on spatial and motor processes when the quality of perceptual evidence extracted is reduced (i.e., by very brief visual exposure). The same stimulus–response bias was not present when the quality of perceptual input was improved (i.e., longer visual exposure). This important finding not only replicates the typical effect of stimulus–response biases, but clearly illustrates that such biases are influenced by perceptual features of stimuli. In sum, perceptual, spatial, and motor processes all interact to influence speeded decision behavior. When the quality of perceptual evidence is reduced, individuals rely more on stimulus–response biases when making quick decisions.

Response time patterns on error trials tell us more about how these respective processes affect decision-making. First, participants made errors faster when the stimuli were presented for longer. This pattern of behavior is consistent with previous implementations of the diffusion model (Ratcliff, Gomez, & McKoon, 2004; Voss, Rothermund, & Voss, 2004). The logic for this pattern is as follows: when a stimulus is relatively easy to identify this results in a larger drift rate and a smaller range between the decision boundaries. This results in two effects. First, participants are less likely to make an error for such a stimulus. Second, since the decision boundaries come closer together, there is a greater probability of the drift crossing the incorrect decision boundary very early in the evidence accumulation process. In the present study, it is assumed that shapes presented for longer (80 ms) are easier to recognize. The participant does not require as much time to make a decision about the stimulus. Consequently, the participant is less likely to make an error on these easier trials and if the participant does make an error it is likely to occur relatively quickly.

Second, errors were generally faster on location-incongruent trials than location-congruent trials. This effect has important theoretical and practical implications for decision-making research. From a theoretical perspective, this pattern lends additional support to the finding that visuospatial mechanisms can interact with motor mechanisms to influence reaction time. However, behavior on correct trials and error trials show opposing effects. When a stimulus is location-congruent with a response key, reaction time speeds up on correct trials but is slowed on error trials. The explanation for faster RTs on correct, location-congruent trials is straightforward. On these trials, visuospatial codes and motor codes do not conflict. Thus, there is no interference present to slow down responding time. The pattern on error trials is less clear. To explain these results, we need to carefully deconstruct the present task. There are two sources of information that influence the motor response: what the stimulus is and where the stimulus is. If we choose a particular response key (“M”) as a reference, there are four possible combinations a target may appear in: (1) correct-stimulus, congruent location, (2) correct-stimulus, incongruent location, (3) incorrect stimulus, congruent location, and (4) incorrect stimulus, incongruent location. In combination 1, both sources of evidence are consistent with the response key. In combinations 2 and 3, one source of evidence is consistent with the response key (stimulus type, stimulus location, respectively) and one source of evidence is inconsistent. In combination 4, neither source of evidence is consistent with the particular response key. Upon closer examination, error RTs showed a tiered pattern across these combinations. Specifically, mean RT for errors was slowest when both sources of evidence were consistent (M = 707 ms). Error RTs were slightly faster when the correct stimulus appeared but in the incongruent location on the screen (M = 697 ms). RTs decreased further when the incorrect stimulus was presented, but in the congruent location with the response key (M = 663 ms). Lastly, error RTs were quickest when the incorrect stimulus was presented in the incongruent location (M = 595 ms). This pattern may suggest decision times on error trials vary depending on the amount of evidence consistent with a particular response. Specifically, as more sources of evidence consistent with a particular response are present, the amount of time for the drift to reach the “error” boundary (Fig. 1) increases—reflecting that the participant needs more time to make a decision about the stimulus.

An alternative explanation for the pattern in error RT is perceptual evidence and response mapping interact on error trials as well, but the manipulations used in the experiment were not strong enough to detect the effect. In the short (20-ms)-duration condition, error RTs were 44 ms faster on incongruent trials than congruent trials. However, in the long-duration (80-ms) condition, error RTs were 86 ms faster on incongruent trials. This may suggest the additional evidence afforded by longer exposure to a stimulus may add additional time to the drift rate when a participant makes an error. This conclusion should be taken with caution, however. As shown in Fig. 5, participants displayed ceiling effects in the 80-ms-duration condition. Thus, the mean RTs on error trials in the 80 ms condition are derived from a relatively small amount of data. Future studies could test this possibility further by including shorter exposure durations, and perhaps including more than two levels of presentation time. A practical implication from the present results is experiments using the keypress modality to make inferences about cognitive functions need to consider patterns on error trials as well as correct trials. It is common practice to omit error trials from RT analyses, but patterns in error trials can reveal more about the processing at play in the respective task.

Table 2 shows the output of the diffusion analysis for each model parameter. The parameters of interest for the present study are the drift rate (v); the boundary separation and starting point (a and z, respectively); and the non-decisional component of the RT (T_er). The drift rate is considerably larger in the long-duration condition (bottom two rows of Table 2), which reflects easier discriminability when the stimulus is presented for longer. This also explains why both correct responses and errors tended to be faster in the long-duration condition. However, the pattern in the drift rates is more complex when we consider the combined effects of stimulus duration and location. On short-duration trials, the drift rate is larger (2.04) when the stimulus is location congruent with the response key than when it is location incongruent (1.28). In contrast, the drift rates are similar between the mapping conditions on long trials. An advantage of the diffusion model is that parameters can be described in terms of specific psychological processes. Recall that the drift rate specifically reflects the decision component (i.e., extraction of perceptual evidence from the stimulus) of the present task, not encoding or motor processes. Thus, we can conclude the larger S–R congruency effects shown on short-duration trials is partially the result of perceptual-decisional processes (reflected in the drift rate). However, we also need to consider the other parameters.

We can measure bias towards one of the decision boundaries by considering the starting point (z) relative to the distance between decision boundaries (a). Focusing on the long-duration condition, z/a = 0.31 on congruent trials and z/a = 0.28 on incongruent trials. This suggests participants display slightly greater responding bias when the shape appeared on the same side as the response key. However, the opposite pattern is shown in the short-duration condition, where z/a = 0.37 on congruent trials and z/a = 0.42 on incongruent trials. This unexpected result suggests that when a stimulus is presented for only 20-ms participants showed greater response bias when the stimulus was displayed on the opposite side from the response key. Additionally, Table 2 shows that non-decision processes (T_er) had a larger effect in the incongruent, short-duration condition relative to the other experimental conditions. This finding is significant since it suggests differences in participant performance across these duration x visibility manipulations may not be solely due to perceptual/cognitive processes, but partially non-decisional processes (i.e., encoding and/or motor processes).

Figure 6 illustrates the observed response quantiles plotted against the model-predicted quantiles for both correct and error trials across the 8 experimental conditions (2 duration x 2 mapping x 2 shape). As can be seen, the model predictions came very close to the observed data. Quantile plots provide a qualitative picture of the model fit; however, the D*M approach obtains parameter estimates by minimizing the Chi-square difference between the observed data distribution and the model distribution for each condition-response pair and then summing the Chi-square values across all condition-response pairs. Results showed a very close fit between the model predictions and the observed data, χ² = 2.89. Overall, the results of the diffusion analysis suggest (1) the diffusion model can adequately account for decisional biases elicited by spatial-response mapping manipulations, and (2) decisional biases are not reflected in a single process. Instead, we see evidence of differences reflected in decisional processes (i.e., drift rate), the starting point of the decision process (z), and non-decisional processes (T_er)—consistent with the interaction hypothesis.

Experiment 2

The findings of Experiment 1 show that S–R congruency effects are modulated by the perceptual quality of evidence afforded by a stimulus. However, presentation duration is not the only feature of a stimulus that may influence the perceptual quality of evidence needed to make an appropriate response, and it should not be assumed that all ways of degrading perceptual evidence elicit the same pattern of decisional biases. A large body of research has shown that stimulus size can produce adverse effects on visual object recognition (Craddock & Lawson, 2009; Graf, 2006; Te Pas et al., 1996; Tomkins, 2021), and stimulus size may also influence S–R congruency biases. For instance, a recent report (Wühr & Richter, 2022) showed S–R congruency effects are modulated by the relative size of shape stimuli, with faster responses to smaller targets on the left and larger targets on the right. At present, however, it is unclear whether stimulus size manipulations specifically affect perceptual, motor, or a combination of both processes in speeded decisions. To examine the generalizabilty of the findings of Experiment 1, Experiment 2 examined the relationship among perceptual, decisional, and response processes further by manipulating stimulus size rather than duration.