A double-blind, placebo-controlled, repeated measures, triple crossover design was employed. Across three visits separated by at least one week, participants were orally administered either cabergoline (1.25mg), amisulpride (400mg), or placebo (sugar pill). Each visit consisted of identical procedures outlined below.

The study was approved by the University of Manchester Research Ethics Committee and was further approved by the UK Health Research Authority for the use of a National Health Service site (Salford Royal Hospital).

Individuals were recruited via university email announcements, public study participation websites, and printed posters around the University of Manchester, UK. In order to avoid recruiting individuals extremely high or low in trait impulsivity, prospective participants completed an online version of the Barratt Impulsiveness Scale (BIS-11, Patton & Stanford, 1995). Participants were then further health-screened in person by a clinician. See Supplementary Materials for further details of the screening procedures, and the power analysis conducted to determine the number of participants for the study.

Thirty-one participants were deemed eligible to participate in the full study (2 left-handed; 16 female; mean age 22.4 years; standard deviation 3.2 years). All participants gave written informed consent and received financial compensation for their participation.

One participant showed an EBR rate more than 1.5 times higher than the interquartile range for the antagonist group and was therefore removed from the EBR analysis. There was a significant linear relationship between dopamine level and EBR; lowest, medium, and highest EBR rates were found in the antagonist, placebo, and agonist conditions, respectively; F(1, 27) = 7.242, p < .05; see Fig. 2. There was no significant nonlinear relationship between EBR and dopamine manipulation (p = .926). This suggests that the dosages of amisulpride and cabergoline administered in this study achieved the hypothesised decrease and increase in striatal dopamine signalling, respectively, due to their action on postsynaptic rather than presynaptic D2 autoreceptors (Frank & O’Reilly, 2006; Maruya et al., 2003).

One participants’ placebo IOR scores were not saved due to computer error. Therefore, their data were not included in further analyses. The remaining twenty-eight participants had a mean incorrect response rate of 3.7% (SD = 1.5). RTs underwent an outlier removal procedure (cf. VanSelst & Jolicoeur, 1994), which removed an average of 2.2% (SD = 0.7) of trials from each participants responses in each visit. All three drug conditions produced slower responses to the cued versus the uncued location.

Using a 2x3 ANOVA, there was a significant main effect of drug on IOR reaction times (cue condition); F(2,54) = 5.92, p < .05, ηp2 = .180. Tests of simple effects demonstrate that both the agonist (M = 344ms; t(27) = 3.30, p = .003, d = .624) and antagonist (M = 342ms; t(27) = 2.65, p = .013, d = .501) significantly decreased RTs compared to placebo (M = 354ms). There was no significant difference between agonist and antagonist RTs (p = .644). The main effect of cue condition was significant; F(1, 27) = 27.70, p < .001, ηp2 = .506. Tests of simple effects revealed that across dopamine manipulations, cued RTs (M = 356ms) were significantly slower than uncued RTs (M = 337ms); t(27) = 5.26, p < .001, d = .994 (i.e. the IOR effect).

The interaction between dopamine manipulation and IOR condition was significant; F(2, 54) = 6.29, p < .05, ηp2 = .189. Planned tests of simple effects determined that cued RTs were significantly faster in the agonist (M = 353; t(27) = 3.94, p = .001, d = .745) and antagonist (M = 349; t(27) = 3.33, p = .003, d = .629) manipulations compared to placebo (M = 366ms). There was no significant difference between agonist and antagonist cued RTs (p = .415). There were no significant differences in uncued RTs between dopamine manipulations: Agonist vs. placebo, p = .025 (Bonferroni adjusted p < .017); antagonist vs.placebo, p = .090; agonist vs.antagonist, p = .950. See Fig. 3.

Cued minus uncued RTs provided a measure of IOR magnitude. There was a significant quadratic relationship between IOR magnitude and drug condition; F(1, 27) = 8.59, p = .007, ηp2 = .241. There was no linear relationship between variables (p = .138). To investigate the nonlinear effect of dopamine manipulation on IOR magnitude, a one-way ANOVA was utilised with three levels: antagonist, agonist, and placebo. The main effect was significant; F(2,54) = 6.29, p = .003, ηp2 = .189. T-tests of simple effects identified that the dopamine agonist (M = 17.9); t(27) = 2.37, p = .025, d = .448; and antagonist (M = 14.9); t(27) = 2.96, p = .006, d = .559; both showed smaller IOR magnitudes compared to placebo (M = 22.9). Therefore, results demonstrate an inverted-U relationship between dopamine level and IOR magnitude. Additional analyses were performed to verify that the observed differences in IOR magnitudes were not caused by drugs generally speeding responses, or confounded by the presence of a speed-accuracy tradeoff, both of which are described in Supplementary Materials.

A one-way ANOVA was performed to test the effect of dopamine manipulation on mean RTs in neutral trials. This was to establish whether there was a significant shift in participant’s baseline RTs which should be accounted for when interpreting the above findings. However, neutral RTs did not differ between placebo (M = 353, SD = 35), agonist (M = 351, SD = 37), and antagonist (M = 352, SD = 37); F(2,54) = .107, p = .899, ηp2 = .004.

The data from three more participants were omitted from ERP analyses (as well as the aforementioned three participants). This was due to the lack of discernible visual-evoked potential in averaged waveforms, defined as a prominent positive peak at posterior electrodes between 70-140ms following stimulus presentation (cf. Di Russo, Martínez, Sereno, Pitzalis, & Hillyard, 2002).

For the remaining twenty-five participants, see Fig. 4 for ERPs obtained for each task condition. No cued minus uncued ERP latency or amplitude showed significant quadratic relationships with dopamine manipulation (see Table 1).

See Table 2 for results of two-way ANOVAs; the significant main effect of IOR condition on N2pc amplitude was further explored using pairwise comparisons. N2pc amplitude was significantly larger in the uncued IOR condition (M = 1.63µV) compared to the cued condition (M = 1.30µV); t(24) = 2.86, p = .009, d = .540; see Fig. 5. This demonstrates that selective attention was stronger for uncued targets compared to cued targets.

No other main effects or interactions reached statistical significance. These null ANOVA results may indicate that the effects of the drugs on IOR may be different for cued versus uncued ERPs, or involve more than a single ERP latency/amplitude.

Further exploratory analyses were guided by behavioural findings, using linear regression analyses to find which ERP(s) best related to notable behavioural effects. Cued and uncued trials were treated separately to account for potential differences in their processing.

As behavioural results demonstrated that the reduction in IOR magnitude were driven by reductions in cued RTs for both drugs, further analyses focused on cued responses; the agonist effect was quantified as mean cued placebo RTs minus cued agonist RTs for each participant. The antagonist effect was quantified as mean cued placebo RTs minus cued antagonist RTs for each participant. Agonist and antagonist effects were also quantified in the same manner for each ERPs latencies and amplitudes.

One participant showed a behavioural antagonist effect more than 1.5 times lower than the interquartile range, and therefore their data were not included in analyses of the antagonist effect. No outliers were identified for the behavioural agonist effect.

The behavioural agonist effect was related to changes in N2pc amplitude; F(1, 24) = 9.709, p = .005, R² = .297. All other ERPs were excluded from the model. This indicated that faster cued RTs in the agonist condition were related to larger N2pc amplitudes relative to placebo (see Fig. 6). No ERPs were significant predictors of the antagonist effect.

Both the D2 agonist and antagonist reduced the IOR effect, relative to placebo. This supports an inverted-U function of striatal dopamine in IOR, with intermediate levels producing greatest effects. Furthermore, we demonstrated the particular involvement of striatal dopamine signalling by using D2R-selective drugs (cabergoline and amisulpride) (Correll et al., 2004; Odin et al., 2006). This is in line with previous findings that either over- or under-signalling of striatal dopamine reduces IOR, such as in individuals with a history of cocaine abuse (Colzato & Hommel, 2009); people with Schizophrenia (Gouzoulis-Mayfrank et al., 2007, 2004); and Parkinson’s disease (Poliakoff et al., 2003). The results are also consistent with Rokem et al. (2012), who found that the D2R agonist bromocriptine increased IOR only in individuals with a lower baseline level of striatal dopamine signalling.

The present study extends previous findings by showing the direct effects of reduced dopamine signalling on IOR, thus demonstrating three points of the inverted-U within the same participant group. Furthermore, we analysed the effects of drugs on cued and uncued trials separately and found that both the D2 agonist and antagonist reduced cued RTs compared to placebo.

There was no statistically significant effect of the drugs on uncued RTs. However, it should be noted that the effect of the agonist on uncued RTs approached statistical significance. Indeed, it is possible that attending towards novel locations might have been faster in the case of uncued RTs, but effects may have been limited by human response speeds (i.e. floor effect) (Liesefeld & Janczyk, 2019). Consequently, it is difficult to conclude if excessive or insufficient striatal dopamine may solely reduce inhibition of previously attended visual locations, or speed the cued and uncued RTs alike.

To understand how both excessive and insufficient dopamine could produce similar behavioral effects, we must consider the underlying neurobiological mechanisms. Dopamine is proposed to have different, even competitive, roles in the striatum versus the prefrontal cortex (Cools & D'Esposito, 2011); prefrontal cortex stimulation via D1Rs stabilises attention and provides resistance to distraction, whereas D2R stimulation encourages cognitive flexibility (Durstewitz, Seamans, & Sejnowski, 2000; Holstein et al., 2011). A study by Riedel et al. (2022) administered L-DOPA, demonstrating a causal influence of dopamine levels on the stability-flexibility trade-off in attentional control during visual search. Other studies have shown that dopamine in the prefrontal cortex plays a crucial role in cognitive control by modulating the motivation to exert control processes (Ott & Nieder, 2019; Cools et al., 2019). The relative balance of prefrontal cortex versus striatal dopamine signaling is mediated by reciprocal frontostriatal connections (Alexander, DeLong, & Strick, 1986; Morris et al., 2016), allowing individuals to both maintain stable task goals whilst preserving the ability to shift attention and update strategies (Cools & D'Esposito, 2011; Cools, Gibbs, Miyakawa, Jagust, & D'Esposito, 2008).

The structural and functional integrity of frontostriatal circuits is crucial for optimal dopaminergic function. Developmental studies show that compulsivity and impulsive traits are linked to attenuated frontostriatal myelination trajectories (Ziegler et al., 2019), while prefrontal and frontostriatal structural variations mediate academic outcomes associated with ADHD symptoms (Chiu et al., 2021). These findings suggest that the circuit-level effects we propose for IOR depend not only on neurochemical balance but also on the structural integrity of frontostriatal pathways. Recent evidence suggests that this balance is particularly important for the subjective effort costs associated with cognitive control, with dopaminergic medications reducing these costs and increasing motivation to exert control (Bogdanov et al., 2022).

IOR is thought to involve the application of an “inhibitory tag” to previously cued locations, following attentional disengagement from that location (Tian et al., 2011). This demarks the location as “already inspected” and therefore not to be prioritised by the attention system (Klein, 2000; Lupiáñez, Klein, & Bartolomeo; 2006). IOR may be considered a mechanism of cognitive flexibility, as it promotes the inspection of novel stimuli in the environment (Colzato et al., 2010; Klein, 2000). Attentional disengagement is thought to be key for the IOR effect (Klein, 2000; Tian et al., 2011), and a lack of disengagement may result in more facilitation effects of the cued location (Colzato et al., 2010). As such, decreased striatal dopamine signalling via the D2R antagonist amisulpride may reduce IOR by biasing the frontostriatal system away from flexibility. A more stable system may hold attention at cued locations, which in turn prevents or delays the formation of the inhibitory tag necessary to produce IOR.

Alternatively, a reduction in striatal signalling may negatively impact inhibitory tag formation, as striatal dopamine signals are thought to be important for the SC to produce such tags (Ikemoto et al., 2015; Klein, 2000). Evidence suggests that the deeper layers of the SC are crucial for attentional modulation and contain specialised functional modules for spatial orienting (Masullo et al., 2019; Ding et al., 2019). The absence of appropriate dopaminergic input from the BG to these specialised SC circuits may bias the attention system toward external sensory cues, causing the sensory element of the stimulus to override IOR. This is in accordance with the general hyper-reflexivity observed in patients with Parkinson's disease (Jackson & Houghton, 1995; Briand et al., 2001; Kingstoneet al., 2002), along with an attenuated spatial and tactile IOR effect (Poliakoff et al., 2003; Possin et al., 2009).

Conversely, an excessive increase in striatal dopamine signalling caused by cabergoline may promote an overly-flexible state, blocking prefrontal cortex activity (Meyer-Lindenberg et al., 2005). The prefrontal cortex may then be unable to maintain inhibitory tags, which would produce faster cued RTs. This would allow for stronger selective attention at the cued location (as supported by our N2pc amplitude findings, discussed in the subsequent section).

The present findings align with evidence that dopamine's effects on cognitive control follow an inverted-U function, where both too little and too much dopamine can impair performance (Badre, 2024; Friedman & Robbins, 2022). Research has shown that dopaminergic medications can improve cognitive control under low cognitive demand conditions (Williams et al., 2020), but effects may depend on baseline dopamine levels and individual differences (Furman et al., 2020). Our results suggest that IOR, as a form of attentional control, may be subject to similar dopaminergic modulation. The inverted-U relationship we observed between striatal dopamine and IOR magnitude may reflect the optimal balance needed for effective cognitive control processes, where intermediate dopamine levels support the most efficient attentional inhibition. This perspective is further supported by research demonstrating that dopamine plays a multi-faceted role in working memory, attentional control, and intelligence through dual-state signaling mechanisms (Matzel & Sauce, 2023).

In summary, we postulate that the drugs cabergoline and amisulpride may shift the balance between responding to novel targets (attention flexibility) versus maintaining inhibitory tags from previous locations (attention stability). In the present study, this may have interfered with inhibitory tag formation and maintenance by the SC and the prefrontal cortex, respectively (thus speeding cued RTs to reduce IOR magnitude). However, in order to clarify the role of dopaminergic frontostriatal connections, future research should directly manipulate dopamine in the prefrontal cortex. Additionally, further research is needed to establish how ERP modulations relate to inhibitory tag formation and the proposed frontostriatal balance in IOR.

While the behavioral findings support this theoretical framework, we also examined electrophysiological measures to gain insight into the underlying neural mechanisms. We found that N2pc amplitudes were greater for uncued versus cued targets, irrespective of dopamine manipulation, showing typical attentional effects on ERPs alongside the behavioural measures (McDonald et al., 2008)

Striatal dopamine manipulations did not significantly affect the difference between cued and uncued N2pc amplitudes or latencies, or that of P1. This indicates that the drugs may have affected cued and uncued target processing differently, and/or affected multiple stages of processing in IOR. Therefore, to account for such complexity, we investigated the potential relationships between how drugs affected RTs and ERPs by analysing their relationship separately for cued and uncued conditions. We found that faster RTs to cued targets in the D2R agonist condition were related to greater N2pc amplitudes. In other words, increased striatal dopamine was related to increased attention to cued targets (Kiss, Van Velzen, & Eimer, 2008).

This finding aligns with more recent evidence that enhanced selective attention processes in IOR can manifest as increased N2pc amplitudes even when behavioral response times are faster (Lin et al., 2020), suggesting that dopamine agonism may optimise the efficiency rather than simply the speed of attentional selection. This tentatively suggests that the D2R agonist reduced inhibition of cued targets, allowing attention to be captured more strongly—thus speeding RTs and reducing overall IOR.

There were no ERP correlates of the behavioural D2R antagonist effect. This could be attributed to several factors, including potential limitations in statistical power for detecting subtle ERP modulations. Other decision-making or motoric processes may have contributed to the behavioural effects that were outside of the measures taken in the present study. Furthermore, the drug may have subtle effects on several processing stages, rather than larger effects on certain stages—making identification of effects more difficult.

In contrast to the null group-level effects, our use of correlational analyses between behavioral and electrophysiological drug effects proved more sensitive than traditional group-level comparisons, revealing that dopamine agonism specifically enhanced the relationship between selective attention (N2pc) and behavioral performance. This approach may be particularly valuable for pharmacological studies where drug effects on different ERP components may vary in magnitude and direction.

In summary, these findings provide some mechanistic insight into the primary behavioural results, indicating that the agonist effect was related to changes in selective attention—whereas the antagonist effect may have influenced other aspects of cognition.

The present findings contribute to emerging research on individual differences in IOR: Bielas et al. (2021) demonstrated that the level of ‘emotional reactivity’ in human temperament can differentially affect IOR magnitude, suggesting that the dopaminergic effects we observed may interact with personality-based differences in attention control.

Our results align with theoretical frameworks that position IOR as a sophisticated information processing mechanism rather than a simple inhibitory effect (Redden et al., 2021; Tsotsos, 2021), with implications extending beyond basic attention research to understanding naturalistic visual behavior and clinical applications.

Previous research has shown that striatal dopamine is important for inhibitory processes (Cools & D'Esposito, 2011), including response inhibition (Bari & Robbins, 2013; Eagle et al., 2008). While correlational evidence suggests striatal involvement in IOR (Poliakoff et al., 2003; Colzato & Hommel, 2009), the present study is the first to directly and pharmacologically manipulate striatal dopamine to examine its causal role in this rapid, reflexive attentional mechanism. Furthermore, some researchers believe that IOR may improve visual search efficiency by reducing unnecessary re-inspections of stimuli (Klein, 2000; Li et al., 2023; Klein, Redden, & Hilchey, 2023). Therefore, the effects of dopamine manipulations on IOR may extend to higher processes such as visual search. It may be reasonable to assume that visual search would be negatively affected by both over- and under-signalling of dopamine, as both drugs reduced RTs to cued locations. Future research is needed to extrapolate these IOR findings to real-world functions such as visual search.

Investigating the specific roles of dopamine signaling in attention may aid in the treatment of attentional disorders. Recent advances in D2R ligands for neuropsychiatric disorders have emphasised the need for more precise targeting of dopaminergic systems (Juza et al., 2023), while research on dopamine receptor expression patterns in ADHD has highlighted the complexity of dopaminergic dysfunction across different attentional disorders (Dum et al., 2022). Our findings suggest that optimal therapeutic approaches may need to consider individual baseline dopamine levels to avoid both under- and over-stimulation of D2 receptor systems.

Recent research has demonstrated that dopaminergic medications can selectively improve cognitive control processes in Parkinson's disease (Cavanagh et al., 2022), and that these effects may depend on the specific cognitive demands of the task (Williams et al., 2020). Multimodal investigations have also revealed complex relationships between dopamine D2/D3Rs, default mode network suppression, and cognitive control in disorders such as cocaine-use disorder (Worhunsky et al., 2021). For instance, dopamine-increasing treatments for Parkinson's disease have unwanted, debilitating effects on attention and impulsivity (Weintraub et al., 2010), which may result from exceeding optimal dopamine levels for cognitive control (Bogdanov et al., 2022). This is thought to be caused by non-selective effects of dopamine drugs on different brain systems, as well as individual differences in patients' dopamine systems (Napier et al., 2015; Weintraub et al., 2006).

The present findings also highlight the need to consider the role of the SC in IOR within the broader context of attention networks. Recent work demonstrating distinct contributions of frontal eye field and SC to covert spatial attention (Bollimunta et al., 2018) suggests that dopaminergic manipulations may differentially affect these parallel attention systems. Given that the SC represents an evolutionarily conserved structure with both general and specialised functions for orienting behavior (Allen et al., 2021; Guillamón-Vivancos et al., 2024), understanding how dopamine modulates SC-mediated IOR may provide insights into fundamental mechanisms of attention that extend beyond human cognition.

Our findings also have implications for understanding attentional disorders across development. Recent research has shown that frontostriatal dysfunction characterises decision-making deficits in both ADHD and obsessive-compulsive disorder (Norman et al., 2018), while dopamine receptor expression patterns vary across different attention-deficit disorders (Dum et al., 2022). The inverted-U relationship we identified suggests that therapeutic interventions for these conditions may need to be carefully titrated to avoid pushing patients from one extreme (too little dopamine) to the other (too much dopamine). This is particularly relevant for pediatric populations, where genetic variations in D2Rs are associated with ADHD symptoms (Moro et al., 2019) and where optimal dopaminergic intervention may differ from adult populations.

In the present study, pharmacologically increasing and decreasing dopamine signalling in the striatum in the same individuals both reduced levels of IOR, supporting an inverted-U relationship between dopamine level and IOR. We propose that these effects emerged from shifts in the balance between attention flexibility and stability, mediated by dopaminergic frontostriatal connections. Such findings have implications for IOR itself, demonstrating the importance of striatal dopamine for attentional inhibition, while also contributing to our understanding of optimal dopaminergic function across multiple cognitive domains. However, there may also be wider implications for the treatment of dopaminergic disorders, particularly the need to consider individual baseline dopamine levels, genetic factors, and circuit-level integrity when designing therapeutic interventions. Future research should investigate the roles of other dopamine receptor subtypes, incorporate polygenic approaches to individual differences, and explore non-pharmacological interventions targeting the same frontostriatal circuits that mediate IOR.