Rapid perceptual decision-making is often key to survival, for example a rat might need to quickly respond to the rustling of leaves to escape a lurking cat. To achieve rapid decisions, the brain needs to process the incoming sensory information, often requiring integration across multiple modalities, to generate an accurate percept and plan and execute the appropriate motor action, often all happening within several hundred milliseconds. The temporal precision of perceptual decisions vary across species; for instance electric fish can resolve 1 μs temporal disparity (Rose and Heiligenberg, 1985; Kawasaki et al., 1988) while echolocating bats can discriminate echo delay changes of 0.4 μs, corresponding to 0.07 mm distance (Moss and Schnitzler, 1989). The resolution at which temporal information is perceived by different species is key to survival and relates to a number of factors including body size and metabolic rate (Healy et al., 2013). For example, fast visual processing with high temporal acuity is necessary for small passerine birds which are high speed flyers that feed from fast moving prey (Boström et al., 2016).

We know that intersensory interactions occur when information from different senses arrives temporally or spatially coincident or near coincident in the brain (Zampini et al., 2005a; Keetels and Vroomen, 2012). The brain must faithfully represent the spatial and temporal features of the external world in patterns of neuronal activity for reliable perception. The temporal features of stimuli such as duration, interval and order of arrival carry crucial information for many aspects of perception (Buonomano and Maass, 2009). For instance, speech (Shannon et al., 1995) and reading (Farmer and Klein, 1995; Hairston et al., 2005; Vandermosten et al., 2010; Wallace and Stevenson, 2014) are disrupted if the temporal information is perturbed. These abnormalities are mainly generated by an extended temporal binding window (Wallace and Stevenson, 2014), for example, in autism (Boer et al., 2013; Stevenson et al., 2014) and schizophrenia (Haß et al., 2017; Stevenson et al., 2017). In the same way, temporal information is critical for how non-human animals, such as frogs, birds and monkeys, can use vocalization for communication (Mauk and Buonomano, 2004).

Discriminating the order of stimuli is a key aspect of temporal processing, which is commonly measured using the well-known temporal order judgment (TOJ) paradigm. In this paradigm, the interval between the onset of two stimuli, also referred to as the stimulus onset asynchrony (SOA), is systematically varied, while requiring the participant to report which side the first of the two stimuli appeared (Boenke et al., 2009), or which modality was presented first (Spence et al., 2003; Zampini et al., 2003b). Using this procedure, one can identify the interval at which the subject can no longer reliably judge the temporal order of stimulus presentation, and this interval is referred to as the point of subjective simultaneity (PSS). TOJ is used to study the temporal processing of both unimodal and multimodal stimuli (Spence et al., 2003; Zampini et al., 2003b; van Eijk et al., 2008; Boenke et al., 2009). Performance depends on the temporal precision of sensory processing within each modality (Keetels and Vroomen, 2012). Much is known about temporal order judgment of multimodal stimuli in humans. For example several factors affect human TOJs including low level factors such as the intensity and duration of stimuli (Boenke et al., 2009), and their spatial disparity (Spence et al., 2003; Zampini et al., 2003b), as well as high level cognitive factors such as attention (Spence et al., 2001), adaptation (Fujisaki et al., 2004; Van der Burg et al., 2013) and the causal relationship between stimuli (Fornaciai and Di Luca, 2020). Another phenomenon which affects the temporal order discrimination, particularly in audiovisual stimuli, is known as multisensory integration (King, 2005; Mégevand et al., 2013; Bedard and Barnett-Cowan, 2016) which refers to integration of two modalities across time. As a result of integration, subjects fail to accurately discriminate the order of audiovisual stimuli or even report them as simultaneous. Audiovisual integration depends on the experimental paradigm (i.e., TOJ versus simultaneity judgment) in which the PSS is estimated (van Eijk et al., 2008; Stevenson and Wallace, 2013; Bedard and Barnett-Cowan, 2016). In order to be perceived as simultaneous, often the visual stimulus has to precede the auditory stimulus (Zampini et al., 2003a; van Eijk et al., 2008; Leone and McCourt, 2015), indicating an inherent bias in humans.

Previous studies have shown that both rats and humans accumulate evidence for perceptual decisions (Brunton et al., 2013; Hanks and Summerfield, 2017). In multisensory context, both species can combine information across auditory and visual modalities to increase accuracy of their choices (Raposo et al., 2012; Sheppard et al., 2013). Integration of sensory evidence across time is a common theme in mammalian decision making (Hyafil et al., 2022). At the behavioral level such integration is manifested in reaction times and accuracy of perceptual decisions in both rats and humans (Hecht et al., 2008; Hirokawa et al., 2011; Gleiss and Kayser, 2012; Hancock et al., 2013). However, these species may use different strategies in resolving perceptual decisions and represent different speed-accuracy trade-offs (Reinagel, 2013b; Shevinsky and Reinagel, 2019). Applying a drift-diffusion model of decision making in the context of visual motion discrimination, a recent study (Nguyen and Reinagel, 2022) has revealed that while variability in the starting point of diffusion process can explain findings in rats, variability in the drift rate can better explain the pattern of human data. In both humans and rats, a small proportion of incorrect choices occur even in easy trials, known as lapse. The rate of lapse seems to be independent of the strength of sensory evidence and mostly reflect lack of attention to stimuli (Wichmann and Hill, 2001; Su et al., 2020). However, a recent study suggested that lapse in rats better explain an exploratory behavior rather than delivering unattended responses (Pisupati et al., 2021). This brief review of evidence clearly reflects similarities and differences in ways that these species resolve perceptual decisions.

Fundamental similarities and differences are also observed in temporal processing in these species. For example, both rats and humans can detect temporal regularities in sound (Celma-Miralles and Toro, 2020). But, in a temporal bisection task humans bisect intervals near their arithmetic mean, whereas rats tend to do the bisection near the geometric mean (Church and Deluty, 1977; Wearden, 1991; Kopec and Brody, 2018). Furthermore, there are reports that humans represent time in a linear manner while rats’ timing behavior is more akin to a logarithmic representation of time (Yi, 2009) (also see Ren et al., 2020 for an opposite view). In delay discounting of reward, which depends on representation of time, the two species act in a similar way but humans seem to be able to wait for much longer durations (Vanderveldt et al., 2016). In addition to interval timing, rodents have also been investigated in order timing. In the first effort for introducing a rodent model of TOJ, Wada and colleagues trained mice to detect the temporal order of vibrotactile stimuli applied to the long whiskers with brief air puffs. They found that mice exhibited lower temporal resolutions compared to humans, and revealed a lateral bias (Wada et al., 2005, 2010). More recent studies in rats (Schormans et al., 2017; Schormans and Allman, 2018, 2019), however, showed that rats could reliably perform TOJ with audiovisual stimuli. The general pattern of findings was similar to humans, but a cross-species investigation was lacking in these studies.

Here, by using the TOJ paradigm in rats and humans, we compare the temporal precision with which these species can detect the order of presentation of simple stimuli across two modalities of vision and audition. To this end, we designed similar audiovisual TOJ paradigms in which rats and humans were required to report which stimulus modality was presented first. Our goal was firstly to better understand how multiple sensory inputs are labeled and reported in their relative temporal occurrence in humans and rats, and secondly to investigate and directly compare the limits of temporal integration across these species. Finally, we performed drift-diffusion modeling on the human and rat data to isolate the contribution of various parameters including evidence accumulation rates, lapse and bias to the sensory decision. This data allows us to better understand how the profile of multisensory processing and labeling generalizes across species. To date, non-invasive methods such as electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) have provided insights about neural underpinnings of multisensory temporal processing in humans (Bushara et al., 2001; Bernasconi et al., 2011; Binder, 2015; Basharat et al., 2018). However, having a comparable rodent model performing in analogous behavioral paradigms is important. This model will provide opportunities for invasive electrophysiological and lesion studies to gain insight to the neuronal computations underlying temporal discrimination or integration of multimodal stimuli.

We implemented a similar TOJ task in rats and humans to test the temporal precision with which these species detect the order of presentation of simple stimuli across two modalities of vision and audition (Figures 1A, B). Each trial contained a visual and an auditory stimulus, and subjects reported which modality was presented first. Here we begin by presentation of results with regard to the temporal discriminability derived from analysis of signal detection theory (Green and Swets, 1966), and then proceed with analysis of reaction time and parameters derived from psychometric function. Finally, we present parameters derived from fitting the drift-diffusion model to the data.

We characterized the temporal precision with which rats and humans could detect the order of presentation of visual and auditory stimuli with a similar TOJ task in both species. We applied centrally-presented pairs of audiovisual stimuli to rats and humans at a range of interleaved SOAs. Both species exhibited high precision in distinguishing the temporal order of audiovisual stimuli with systematic similarities and differences in their parameters of performance. By applying principles of signal detection theory, we found that temporal discriminability (d’) systematically decreased by decreasing the SOA in both species. However, rats showed greater sensitivity at short SOAs (shorter than 100 ms) whereas humans showed greater performance at long SOAs (longer than 100 ms) (Figure 1C). We found that the rats’ higher lapse rate compared to humans could contribute to this profile of performance. A drift diffusion modeling further confirmed that rat’s superior performance at short SOAs was possibly due to faster collapsing decision bounds rather than higher drift rates. Such differences in performance profile may also reveal limitations of each species in detection of temporal mismatch between bimodal stimuli.

It is known that the resolution at which temporal information is perceived by different species depends on their body size and metabolic rate (Healy et al., 2013). For example, in a flicker fusion frequency task small passerines birds distinguished flickering stimuli at temporal precision of around 130 Hz, which is substantially higher than other vertebrates including humans (Boström et al., 2016). An organism’s sensory limitations impose critical constraints on how the organism can interact with its environment. In the current study, better temporal discriminability and accuracy in short SOAs in rats compared to humans support this assumption. On the other hand, lower accuracy in long SOAs in rats compared to humans might reflect the fact that humans primarily focused on accuracy rather than speed whereas rats maximized their reward gain not necessarily via increasing accuracy. For the current experiment, correct responses in rats were accompanied by a sucrose reward, therefore, incentive anticipation of reward probably may have promoted unattended impulsive responses (Donnelly et al., 2015; Marshall and Kirkpatrick, 2016) in a proportion of easy trials (i.e., long SOAs; greater than 100 ms).

In addition to accuracy, we also measured reaction time across SOAs and found that overall rats performed the task with shorter reaction times compared to humans, and this was despite the fact that rats needed to reach the lick ports rather than merely pressing a button. Shorter reaction time in rats may indicate that their strategy in performing the task is different from that of humans. It has been shown that in difficult choices, animals may prioritize responding more quickly when no punishment is expected (Chittka et al., 2009). For encouraging animals to reduce speed and increase accuracy one needs to manipulate task parameters including reward or punishments as reinforcers. For example bees sacrifice speed in favor of accuracy when quinine penalties are applied after error responses (Chittka et al., 2003). In our study, incorrect responses accompanied no reward but also no explicit punishment. It is therefore not surprising that rats focused on speed rather than accuracy in order to maximize reward gain. Another difference between the two species was in the dynamics of speed accuracy trade-off. Generally, the speed accuracy trade-off should be optimized given the specific demands of the environment. The speed accuracy trade-off has been studied for over a century in different species including house-hunting ants (Stroeymeyt et al., 2010), bees (Chittka et al., 2003), mice (Rinberg et al., 2006), rats (Kaneko et al., 2006; Reinagel, 2013a; Mendonça et al., 2020), and human (Herz et al., 2017) and non-human primates (Heitz and Schall, 2012; Hanks et al., 2014). In our study, rats performed the task with faster responses compared to humans. However, such fast responding was even more pronounced in shorter SOAs (i.e., difficult choices) whereas, in line with previous studies, in humans more difficult choices resulted in longer decision times (Figure 1D). For example, in a visual task involving detection of motion direction, increasing difficulty by decreasing fractions of coherently moving dots increases reaction time in monkeys (Roitman and Shadlen, 2002) and humans (Palmer et al., 2005). However, rats revealed a different strategy in performing motion discrimination (Shevinsky and Reinagel, 2019) and visual image discrimination (Reinagel, 2013a) tasks. Shevinsky and Reinagel (2019) designed an analogous motion discrimination task in rats and humans and in line with the current study they observed that while humans had longer reaction time at lower accuracy, rats exhibited shorter reaction time at lower accuracy (Shevinsky and Reinagel, 2019). On the other hand, in a tactile detection task involving whisker inputs, rats increase their sampling time in difficult trials to obtain more information about whisker deflection (McDonald et al., 2014). It is therefore not entirely clear whether the differences in speed accuracy trade-off reflect fundamental differences across species or are partially determined by sensory modality or task parameters.

Another difference between rodents and humans is the temporal window in which they accumulate evidence for making perceptual decisions. For example, rats need about 300 ms to decide between a mixture of two odors while only 35 ms more time is enough for making decisions in difficult trials (Uchida and Mainen, 2003). This time is reported as long as 80 ms in mice (Abraham et al., 2004). These results indicate rodents need a relatively small window of integration compared to what humans or monkeys need to accumulate evidence. In rats, the interplay between speed and accuracy also depends on the task. For example, in both olfactory identification (single pure odor) and olfactory categorization (mixture of two pure odor was presented) tasks, reaction time increased with task difficulty but with smaller trends in the categorization task (Mendonça et al., 2020). This indicates that categorization benefits less from temporal integration. Of great relevance to the context of the current study is TOJ of multisensory stimuli, which is rarely investigated in rats (Schormans et al., 2017; Schormans and Allman, 2018). As a common phenomenon in multisensory contexts, integration of modalities may improve target detection, but it interferes with tracking the order information (Miller and Schwarz, 2006; Mégevand et al., 2013). Indeed, integration of information across modalities hinder discrimination of their order in TOJ and therefore increases reaction time in difficult choices (i.e., short SOAs), although it is not a common theme across studies of this kind [e.g., see Leone and McCourt, 2015]. Also, a clear distinction exists between TOJ and simultaneity judgment tasks (Love et al., 2013; Bedard and Barnett-Cowan, 2016; Basharat et al., 2018). In contrast to TOJ, integration of modalities in simultaneity judgment, is in line with perceiving them as simultaneous, particularly in short SOAs, which come up with a reduction in reaction time (Matthews et al., 2016; Simon and Wallace, 2018). We found that in humans, reaction time increased with decreasing SOA, whereas an opposite pattern was observed in rats. Whether this discrepancy depends on the strategy that the two species used for discrimination of stimuli or the way that they deal with multisensory integration remain to be addressed in future studies. We also found shorter reaction times in correct visual-first choices in both species. This is not consistent with the observation that unisensory reaction time is faster in auditory than visual stimuli (Leone and McCourt, 2015; Bella and Silvestri, 2017; Harrar et al., 2017), and the fact that processing time of auditory modality is shorter than that of the visual modality. Whether a faster reaction in visual choices depends on the task demand and is specific to TOJ needs further investigation.

In this study, we applied a cumulative Gaussian function to data to quantify psychophysical aspects of human and rat performance, including bias or PSS, slope and lapse rate. PSS indicates the temporal delay at which the two stimuli are perceived as simultaneous: i.e., where the probabilities of visual-first and auditory-first choices are equal. We found a PSS of −4 ms in rats, but this was not significantly different from zero indicating that they were unbiased to visual and auditory stimuli. However, a recent study showed an average PSS of −8.8 ms across seven rats (Schormans et al., 2017). In humans the average PSS was +19 ms however, again this was not statistically different from zero. Despite the numerical difference, PSS was not significantly different between humans and rats. From previous studies we know that the value and sign (negative or positive) of PSS can vary with the intensity (Schormans and Allman, 2018), or duration (Boenke et al., 2009) of stimuli as well as with task conditions (Zampini et al., 2005b). Another key finding in psychometric parameters was the slope of the fitted sigmoid as a measure for temporal sensitivity. In our study, rats exhibited a higher slope and consequently higher temporal sensitivity at short SOAs compared to humans. Also, rats revealed a higher lapse rate compared to humans which predominantly was derived from a large visual lapse in rats (Figure 2C). In similar studies in rats (Schormans et al., 2017; Schormans and Allman, 2018, 2019), there is no report of lapse, so it remains to be clear in future studies whether this is an inherent bias in rats or depends on the experimental paradigm. The higher lapse rate along with shorter reaction time in rats indicates that two species have adopted different strategies in resolving the task. Consistent with our findings, a higher lapse rate has been reported in rats compared to humans in a motion discrimination task (Shevinsky and Reinagel, 2019) reflecting a species difference in speed-accuracy strategy so that rats’ reaction time strongly affects by elapsed time during a trial rather than strength of evidence (Reinagel, 2013b). When discussing drift diffusion model, we further elaborate the species differences in terms of decision strategy.

In addition to measuring accuracy and speed across the range of SOAs, we also applied a drift-diffusion model to investigate differences and similarities between decision process in humans and rats. As discussed earlier, making fast decisions is expected to come at the expense of reduced accuracy. A number of computational models including the stochastic accumulator models (Usher and McClelland, 2001; Smith and Ratcliff, 2004; Tavares et al., 2017; Roxin, 2019; Shinn et al., 2020) are used to quantify speed accuracy trade-off. These models allow us to isolate fundamental aspects of a decision making process, including the rate of sensory information, the arrival to a decision threshold for making choices, and the lapses and biases that influence sensory decisions. We applied the same model to our human and rat data to reveal fundamental similarities and differences by which these species make decisions regarding the temporal order of stimuli. To this aim, a variant of drift-diffusion model which considers collapsing boundaries for decision over time was fitted to the data. This model can capture the data in rats (Mendonça et al., 2020) and monkeys (Hawkins et al., 2015) when learning and training shape the decisions. Another reason for using this model was that we used a short range of asynchronies and therefore evidence with regard to stimuli’ order can be degraded over time, which corresponds to collapsing decision boundaries across time (Hanks et al., 2014; Voskuilen et al., 2016; Malhotra et al., 2018). This model assumes that in a given trial, the requirement of evidence for making decisions decreases as a function of time. Also, it has been shown that a cost exists for spending time on tasks in human and non-human primates because accumulation of information is a demanding process (Drugowitsch et al., 2012) and therefore collapsing boundaries prevent consuming limited cognitive resources. We found a similar picture, particularly in rats’ data, so that they revealed lower leakage, performed the task with shorter reaction times and outperformed humans in short SOAs. Unlike the aforementioned primate motion discrimination paradigm (Roitman and Shadlen, 2002; Hanks et al., 2014), where sensory evidence continuously flows in, here the information is provided in a single time at the onset of the trial. Therefore, any delays in the decision can potentially corrupt the evidence through leakage (Drugowitsch et al., 2012). Faster and more accurate choices in short SOAs in rats is also consistent with faster collapsing boundaries in rats. The model thus explained why despite having lower decision thresholds and higher lapse rates, rats maintained a higher level of performance compared to human participants, particularly in short SOAs.

To allow comparison between species, we designed an analogous TOJ task for humans and rats. However, important differences remain between the two paradigms and the training schedules, which should be taken into consideration when interpreting the findings. Humans could be instructed as to the requirements of the perceptual task, but rats needed an extensive training protocol to shape their behavior. The rats were rewarded for correct choices and therefore had a different level of motivation compared to human participants. The sucrose reward also produced an immediate feedback to the rats regarding the choice on every trial. We therefore cannot rule out the potential impact of feedback or difference in the duration of training/familiarization on the observed species differences in the current study. Additionally, the type of response required differed between the two species: rats responded by moving their body towards and licking from the reward spout, whereas humans performed a manual response. Another difference was the application of 15 ms SOA in rats but not in humans. The range of SOAs was determined based on performances. Human participants found temporal discrimination at the 25 ms SOA difficult. Indeed, only half of the human participants performed better than chance at this SOA, whereas all rats performed better than chance for the same SOA. We therefore extended the SOA for rats to also include the 15 ms interval which was the smallest that we could reliably implement. Finally, mapping auditory and visual modalities to the left/right side was equally counterbalanced across rats, whereas, counterbalancing was not the case in human subjects. All subjects (right-handed) in the visual-leading and auditory-leading trials responded to the left and right sides, respectively. However, we did not observe any systematic differences in reaction times, and lack of bias in the PSS implies that overall performance was not influenced by nonspecific factors such as handedness or any potential intrinsic left/right bias in subjects. Only a modality-dependent bias was observed for reaction time, such that the average reaction time was shorter in the visual choices. Because visual choices were mapped to the left hand, therefore shorter response times in the left/visual choices (compared to right-auditory choices) suggests the presence of a modality effect rather than a handedness bias which would have been expected to favor the right choices in right-handed individuals. Future studies could adjust the human and rat paradigms to remove some of the methodological differences here including the extent of training and the presence of feedback. These studies could also record signals in the form of EEG from humans and neural activity in behaving rats undertaking this task to investigate the neuronal mechanisms that underlie perceptual decisions in such multimodal contexts.

By implementing a similar psychophysical paradigm in rats and humans, we found that both species could reliably distinguish the temporal order of audio-visual stimuli based on the modality label. Rats showed superior performance at difficult trials, with SOAs shorter than 100 ms, whereas humans showed superior performance at easier trials, with SOAs greater than 100 ms. Rats exhibited a high temporal precision by performing better than chance even at the 15 ms SOA. Overall, rats produced faster responses compared to humans and did not increase their reaction time as difficulty increased. Despite their higher temporal precision, rats’ overall performance was hampered due to a high level of lapse rate and this was most evident at the easy trials with long SOAs. Despite a few similarities, our findings indicate that rats and humans apply different strategies in terms of speed and accuracy in resolving perceptual decisions when the difficulty of decisions varies across trials.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Ethics Committee of Kerman University of Medical Sciences. The patients/participants provided their written informed consent to participate in this study.

FM, MA, and EA designed the initial project. FM performed the experiments. All authors contributed to the data analysis and interpretation. FM, SG, and EA designed and built the experimental apparatus and wrote code to control the stimuli and monitor the behavior. MT performed the initial modeling. FM and EA drafted the first version of the manuscript. All authors contributed to the subsequent revisions.