Parameters of motion vision in low light in the hawkmoth Manduca sexta

Nocturnal insects display a variety of spectacular visually guided behaviors despite the dim light conditions in which they operate. Examples of these include the navigation of dung beetles using celestial cues (el Jundi et al., 2015), the ability of carpenter bees to learn landmarks at night (Somanathan et al., 2008) and color vision in extremely low light conditions in a nocturnal hawkmoth (Kelber et al., 2002). Optical and physiological adaptations enable nocturnal insects to successfully operate in their temporal niches (Laughlin and Weckström, 1993; Warrant, 1999; Stöckl et al., 2016a). One well-known physiological adaption underlying low light vision in nocturnal insects is the superposition compound eye, which increases the eye's sensitivity to light by integrating light spatiotemporally (Warrant, 1999; Theobald et al., 2006; Warrant and Dacke, 2011). Superposition compound eyes increase light capture by pooling signals from neighboring ommatidia, but in doing so, they impose a limit on detectable contrast and spatial and temporal frequencies (Cronin et al., 2014).

The hawkmoth Manduca sexta has been reported to be active from twilight to darkness (Sasaki and Riddiford, 1984; Sponberg et al., 2015; Broadhead et al., 2017; Stöckl et al., 2017). Despite the possible limitations imposed by these low light conditions, M. sexta is known to perform precise flight maneuvers in dim light (Willis and Arbas, 1991; Willmott and Ellington, 1997; Sponberg et al., 2015). Manduca sexta displays characteristic odor tracking behavior with counterturns finding food and mates (Willis and Arbas, 1991; Raguso and Willis, 2003). Mating and nectar foraging behavior is activated and driven by odor information acquired by the olfactory system, and input from the visual system is required for flight control (Raguso and Willis, 2005; Balkenius and Dacke, 2010; Willis et al., 2011). Odor tracking requires the moths to perform flight maneuvers in dim light to locate and lock on to the odor plume, orient to and move into the wind direction, and orient towards specific objects and avoid obstacles. Nectar feeding behavior requires the moths to hover and hold station in front of flowers while they are buffeted by wind and the down draft from their own flapping wings (Farina et al., 1995; Sprayberry, 2009; Sponberg et al., 2015; Stöckl et al., 2017). In any odor-modulated flight behavior performed by these moths, information gathered and processed through the visual system is thus of critical importance for flight stability and control.

Much of basic flight stability and control relies on the moths' ability to detect and respond to wide-field visual motion (Borst and Haag, 2007). As the animal moves around in the environment, the visual system receives a constantly changing optic flow field (Lee, 1980; Koenderink and van Doorn, 1987). This self-induced wide-field motion is used by both nocturnal and diurnal insects to stabilize their flight and control flight maneuvers (David, 1979; Krapp et al., 2001; Barron and Srinivasan, 2006; Dyhr and Higgins, 2010; Shafir and Barron, 2010; Theobald et al., 2010; Baird et al., 2011; Rutkowski et al., 2011). While the wide-field motion vision of diurnal insects, especially flies, is well studied (Egelhaaf et al., 2002; Borst et al., 2010), the same depth of knowledge of the wide-field motion vision of nocturnal insects is emerging slowly (Warrant and Dacke, 2011).

It is thought that the wide-field motion vision computation is primarily done by the lobula plate tangential cells (LPTCs) of the lobula complex located in the optic lobe of the insect brain. The LPTCs are sensitive to visual motion and have a preferred and non-preferred direction of motion; that is, they respond selectively to the motion direction of visual stimuli (Borst et al., 2010). The response characteristics of the LPTCs to wide-field motion stimuli across a range of dim light conditions of hawkmoths and M. sexta in particular have been studied recently (Theobald et al., 2009; Stöckl et al., 2016a). The approach of using optomotor responses measured from tethered moths have been previously used to study the visually aided flight control of crepuscular moths (Dyhr et al., 2013; Windsor et al., 2014). However, the effect of luminance, a defining characteristic of their environment and activity period, on the steering behavioral response of M. sexta to wide-field motion has not been studied systematically.

We measured the attempted steering response of tethered M. sexta to experimentally manipulated changes in wide-field visual motion in a range of dim light conditions, from daylight to a dark moonless night. Under these light levels, steering responses were characterized for a range of contrast levels, spatial frequencies, and temporal frequencies. We found that the attempts of M. sexta to turn in response to wide-field motion are tuned to its temporal niche, sunset to moonlight, and the preferred spatial frequency decreases as the luminance level decreases. The preferred temporal frequency and the time to respond to a change in motion direction did not vary significantly across the luminance level. Together these results suggest that M. sexta can adapt to a wide range of light levels, and that, at least for the type of steering responses supported by wide-field visual motion, spatial resolution may not be as important as speed of response in the low-light conditions in which these animals normally operate.

We used 3- to 5-day-old adult Manduca sexta (Linnaeus 1763) moths reared from our laboratory colony and maintained on a 14 h:10 h light:dark cycle. The experiments were conducted during the dark phase of the light:dark cycle.

Moths were removed from the environmental chamber an hour before their subjective sunset. They were immobilized by placing them on ice for 10–15 min under room lighting conditions. Before tethering, all legs were surgically removed and the scales on the thoracic sternum were removed. A 2-cm-long, 18-gauge stainless-steel tube with a flattened end was attached to the moth between the middle and hindlegs using cyanoacrylate glue. The pitch angle of the longitudinal body axis of the tethered moth with respect to the horizon was set to ∼30 deg, similar to the observed pitch angle of M. sexta flying freely in a laboratory wind tunnel (Rutkowski et al. 2009).

A mini projector (P4X Pico Projector, AAXA Technologies, Irvine, CA, USA) was used to project the visual stimuli onto a dome-shaped rear-projection screen (diameter, 67 cm; height, 30 cm) (Fig. 1A). Owing to the space limitation in our experimental set-up, we used a mirror (76×60 cm) to achieve the desired image size from the projector. The brightness of the projected images was fairly uniform across the dome except at the very edges of the dome, where we measured an approximate 6% reduction in the projected luminance value. Custom written MATLAB code using Psychophysics Toolbox-3 (Brainard, 1997; Pelli, 1997; Kleiner et al., 2007) was used to generate sinusoidal gratings, control stimulus presentation, and trigger the data acquisition system (Neuralynx, Bozeman, MT, USA). The frame rate of the projection was approximately 60 Hz.

We presented gratings at light conditions ranging from daylight to starlight. These conditions were achieved with neutral density filters (Lee Filters, Burbank, CA, USA). The light levels were then measured with a light meter (ILT1700 Research Radiometer, International Light Technologies, Peabody, MA, USA) by placing it in front of the concave side of the dome. We then added neutral density filters with different stops to achieve the desired light conditions. The following luminance levels (cd m⁻²) were used in this study: 70, 10, 1, 0.1, 0.01, 0.001 and 0.0001 cd m⁻². The tethered flight arena was isolated from sources of light (i.e. computer monitor, etc.) by surrounding it with two layers of thick black fabric curtain. In addition, care was taken to minimize overall light level in the recording room.

Each trial was 6 s long and consisted of a sinusoidal grating moving to the moth's right for 3 s, then to the left for the remaining 3 s (Fig. 1B; Movie 1). We varied the parameters of the grating, and each stimulus was a unique combination of the following: spatial frequencies (cycles deg⁻¹): 0.0075, 0.015, 0.03, 0.045, 0.06 and 0.075; temporal frequencies (Hz): 1.5, 3, 4.5, 6, 7.5 and 9; and contrast values (%): 1, 5, 10, 15 and 20. The trials were randomized and repeated three times. Stimulus presentation was followed by a 2 s inter-trial interval, during which the screen was a medium grey. At the end of each recording session the screen was set to black. The light level of the stimulus was changed and presented in descending order to imitate the moth's daylight cycle. At each decrease in luminance level, the moths were given 10–15 min to dark adapt their compound eyes. The recording session for each light level lasted ∼72 min, sufficient time to adapt to the light level.

Two white squares were also projected onto the bottom of the monitor screen along with the gratings. These squares changed from black to white at the beginning of each trial and when the pattern movement direction changed. Photodiodes positioned in front of these squares on the monitor screen detected the changes in brightness of the square which was recorded synchronously with the behavioral responses for offline analysis.

The tethered moth was placed inside the dome at a distance of 135 mm from dome's center of curvature. The set-up included an overhead camera mounted perpendicularly to the moth, capturing images at 60 frames s^–1 connected to a custom-developed (Python and OpenCV) online image tracking system to obtain head position and wing steering information. This set the temporal resolution of steering response to ∼17 ms. The dorsal surface of the moth was illuminated evenly using an array of infrared LEDs to ensure accurate measurements of head and wing positions. The wing steering information was extracted by measuring the angle between the longitudinal body axis and the tracked leading edges of right and left forewings. The intersection of scutum and scutellum sutures was used as a reference point to mark the longitudinal body axis midpoint. The head position was extracted by tracking the location of center of mass of the tracked head in a given frame. The digitized head position and inter-wing angles were converted to analog signals and recorded. The steering behavior, trial duration and motion direction signals were digitized at 1000 Hz using a Neuralynx data acquisition system for offline analysis.

We used custom written MATLAB scripts for data analysis. The digitized head turning and wing steering values in volts were used for the analysis. To capture the magnitude of the response during the trials, we used area under the curve measurements from the head turning and wing steering response (Fig. 1C). We observed some aliasing response, i.e. the moth's optomotor response was in the reverse direction of the grating motion, at low temporal frequency (1.5 Hz). The aliasing responses were identified using the slope value of the response phase when stimuli switched direction (Fig. 2) and excluded from further analysis. The strength of the steering response to the stimuli was calculated in terms of the area under the curve for each steering response during the stimulus presentation. The area under the curve values were used to construct contour maps of spatiotemporal optomotor responses (Fig. 2).

where a and d are the minimum and maximum asymptote, respectively, b is Hill's slope of the curve, and c₅₀ is the half-maximal effective light level.

The spatiotemporal frequencies that produced maximal response in each light condition were identified from the area under the curve values. The area under the curve values are assembled into a 6×6 matrix (spatiotemporal matrix), then smoothed using Gaussian filters. The spatiotemporal frequencies that elicited maximal response were identified for each light condition. The optomotor response obtained in each trial was the result of a unique combination of the following visual stimulus parameters: contrast, spatial and temporal frequency and luminance. We used principal component analysis (MATLAB 9.2) to examine the effects of luminance on both the spatial and temporal frequency response. We also have compiled a response profile by taking the average across the dataset (Figs S1 and S2).

The area under the curve value of the trials were sorted based on the spatial frequency and the contrast value, ignoring the temporal frequency of the trial. We were left with the optomotor responses to all the spatial frequencies presented at different contrast values (1 to 20%). We used principal component analysis to obtain a representative response to all the spatial frequencies for each luminance level. The process was repeated for temporal frequency analysis and the data were sorted based on the temporal frequency ignoring the spatial frequency. The head turning and wing steering response data for each moth (N=15) were analysed separately. The principal component that accounted for >40% variation in the dataset was used to represent the data. Based on this criterion we used the first principal component to represent the level of optomotor response of the moth to the given spatial and temporal frequencies. Spatial and temporal frequency profiles were constructed using first principal component scores. The data were further subdivided into groups based on contrast and averaged to find preferred spatial and temporal category in each group. The response profiles were compiled using both the second and third principal component scores, which accounted for up to 27% of the variability in the results, for both the temporal and spatial frequencies and are presented in the supplementary material (Figs S3 and S4).

To estimate the effect of luminance level on the moth's ability to follow wide-field motion, we measured onset latency and response speed. Onset latency is defined as the time taken by the moth to initiate an optomotor response after the onset of the stimulus (Fig. 1D). We used the 95% confidence interval of the mean response as a threshold to mark the beginning of the response. This captures the time taken by the moth to process the visual stimuli. Response speed is the time taken by the moth to turn to follow the grating when the grating switched directions of motion. The time from the peak response that occurred just before the change of motion direction to when the signal crossed the baseline is defined as response time (Fig. 1D). The response time was measured for each temporal frequency. Responses that were less than 5% of peak response were removed from this analysis. Based on this criterion the responses for the light level below 0.01 cd m⁻², i.e. moonlight level, were excluded from the analysis.

To see how the relationship between the temporal frequency of the stimulus and the resulting response was affected by the contrast and light levels, we fitted a linear function to the response time data for the temporal frequencies to obtain slope values. We then looked at these slope values obtained for each light level at different contrast values. In addition to these temporal measurements we also measured switching latency, i.e. the time between onset of a new motion direction and when the response begins to roll off. The roll off was estimated by measuring the sign change of the slope of the response curve. This indicates the time taken by the moth to detect and respond to the change of motion direction of the stimuli. The data for this analysis are provided in Fig. S5.

Tethered flying adult M. sexta moths were presented with moving wide-field stimuli at a range of light levels from 70 to 0.0001 cd m⁻². Attempts to turn to follow the wide-field pattern movement (i.e. the optomotor response) were used as an indication of the moth's ability to detect and follow the stimuli. A ‘map’ of the moths' responses to the contrast values and spatial and temporal frequencies of the wide-field motion stimuli at different light levels was generated by systematically changing these parameters of the sinusoidal gratings. The area under the curve of the steering response was used as an indicator of the strength of moths' responses to the corresponding stimulus (Fig. 1; see Materials and Methods for details on analysis).

The overall optomotor response profile of both the head turning (Fig. 2A) and wing steering (Fig. 2B) were similar. Moths showed neither head turning nor wing steering to the moving patterns presented with contrast value of 1%. Also, we did not observe an optomotor response of any kind at the light levels below moonlight (0.01 cd m⁻²). The optomotor response started to emerge as the contrast increased from 1 to 5%, and when the stimuli were presented at moonlight level (0.01 cd m⁻²). There was a weak aliasing response when the stimulus contrast was high and the light level was above 0.1 cd m⁻² (Fig. 2).

The response profile for both head tracking (Fig. 3A) and wing steering (Fig. 3B) was constructed using peak response produced by moths for sinusoidal gratings with different contrast presented at a range of light conditions. Moving wide-field motion cues presented with 1% contrast did not elicit optomotor responses at any of the experimental light levels (Fig. 3A,B). These data also suggest that 5% contrast is near the lower limit of detectable contrast under our experimental conditions. To estimate half-maximal effective light conditions, we fitted data, except for 1% contrast, using a four-parameter logistic linear regression model (continuous red line in Fig. 3A and B; head position data R²=0.87; wing steering data R²=0.85). The half-maximal values for head tracking and wing steering were 0.015 and 0.018 cd m⁻², respectively. Looking at the linear phase of the logistic regression function it could be suggested that the visual system of M. sexta is adapted to respond best between the light conditions 0.1 and 0.001 cd m⁻² (Fig. 3A,B). To find the effective light level for each contrast used in our experiment, we estimated the half-maximal effective light level, c₅₀, the luminance value halfway between the lowest and the highest optomotor response (Fig. 3C). The minimum contrast that could be perceived increased as the light level reduced. The lower limit for perceivable contrast is between 5 and 1% in twilight (Fig. 3C).

The moths' preferred temporal frequency remained essentially the same through all of the light levels presented (Fig. 4). The response amplitude was reduced at light level 0.01 cd m⁻² and the preferred temporal frequency remained at 4.5 Hz (Fig. 4A: two-way unbalanced ANOVA: within the group ‘light level’, F_6,582=229.6815, P<0.001; within the group ‘temporal frequency’, F_5,582=205.0298, P<0.001; between groups, F_30,582=16.8282, P<0.001; Tukey’s post hoc comparison of groups; Fig. 4B: two-way unbalanced ANOVA: within the group ‘light level’, F_6,582=118.2876, P<0.001; within the group ‘temporal frequency’, F_5,582=116.1845, P<0.001; between groups, F_30,582=9.5525, P<0.001; Tukey’s post hoc comparison of groups). The peak response for each light condition and contrast shows that 4.5 Hz remains the preferred temporal frequency throughout our experimental conditions despite the changing contrast levels (Fig. 4C). The data for the second and third principal component (PC) scores are provided in Fig. S3. The PC2 score of wing steering data showed a peak response to 3 Hz temporal frequency only for light levels above 0.1 cd m⁻² (Fig. S3C). The PC2 score for the head turning and PC3 scores for both the head turning and wing steering data did not show any response to different temporal frequencies (Fig. S3B,D).

The spatial frequency response profiles remained the same for light conditions 70 to 1 cd m⁻² (Fig. 5A,B). As the available light decreased below 1 cd m⁻² the moths' response became weaker and their ability to follow wide-field motion cues with higher spatial frequency decreased (Fig. 5A: two-way unbalanced ANOVA: within the group ‘light level’, F_6,582=319.2043, P<0.001; within the group ‘spatial frequency’, F_5,582=86.7616, P<0.001; between groups, F_30,582=10.0267, P<0.001; Tukey’s post hoc comparison of groups; Fig. 5B: two-way unbalanced ANOVA: within the group ‘light level’, F_6,582=142.2697, P<0.001; within the group ‘spatial frequency’, F_5,582=63.833, P<0.001; between groups, F_30,582=8.3603, P<0.001; Tukey’s post hoc comparison of groups). The peak response for each light condition and contrast was extracted from the response profile. As the light condition dropped below 0.1 cd m⁻² the moths' responses shifted to lower spatial frequency (Fig. 5C; see Materials and Methods). The PC2 scores of both head turning and wing steering data showed response to lower spatial frequencies at light levels higher than 0.01 cd m⁻² (Fig. S4A,C). The PC3 score did not capture any response to the changing spatial frequency (Fig. S5B,D).

We measured the moths' optomotor response latency to wide-field motion cues at different light levels. For the purposes of this analysis, the time between the stimulus onset and initiation of the optomotor response is defined as onset latency (Fig. 1D). The trials were sorted based on temporal frequencies of the stimuli presented. The time taken by the moth to initiate a turn is inversely proportional to the temporal frequency at given luminance levels (Fig. 6A–E; R² values obtained for shown data: 0.23, 0.48, 0.43, 0.24 and 0.001; P<0.05 for the light levels 70 to 0.1 cd m⁻² and P=0.86 for the light level 0.01 cd m⁻²). To see how this response was affected by both contrast and luminance levels, we fitted a linear function to the response time data and used slope values to compare these parameters (Fig. 6A–E; R² values for shown data: 0.23, 0.48, 0.43, 0.25 and 0.001). We noticed that the slope value increased slightly at both twilight and moonlight conditions except for responses to patterns with 20% contrast. However, this change was not statistically significant (two-way unbalanced ANOVA: within the group ‘light level’, F_4,260=1.77, P=0.1353; within the group ‘contrast’, F_3,260=4.93, P=0.0024; between groups, F_12,260=1.08, P=0.3761).

Finally, we measured the moths' attempted turning response speeds to wide-field motion cues at different light levels. The time between the maximum head deviation of the moth occurring immediately after the stimuli switched direction, and time to reach the center (i.e. no detectable lateral bias), was measured and referred to as ‘response speed’ (Fig. 1D). The trials were sorted based on temporal frequencies of the stimuli presented. The response speed is inversely proportional to the temporal frequency at given luminance levels (Fig. 7A–E). To see how this response was affected by both contrast and luminance levels, we fitted a linear function to the response time data and used slope values to compare these parameters (Fig. 7A–E; R² values obtained for shown data: 0.30, 0.47, 0.39, 0.17 and 0.01; P<0.05 for the light levels 70 to 0.1 cd m⁻² and P=0.68 for the light level 0.01 cd m⁻²). We noticed a small increase in slope values for both starlight and moonlight conditions (Fig. 7F), which was not statistically significant except for 5% contrast at the moonlight condition (two-way unbalanced ANOVA: within the group ‘light level’, F_4,267=6.87, P=0; within the group ‘contrast’, F_3,267=3.09, P=0.027; between groups, F_12,267=1.68, P=0.719; Tukey’s post hoc comparison of groups).

In this study, we show how changing light levels shape the optomotor response of tethered flying M. sexta moths. In doing so, we provide for the first time the basic performance envelope for the optomotor response of M. sexta for a range of contrast and spatial and temporal frequencies. These results indicate that in M. sexta, this visually driven behavioral response is tuned to the moths' crepuscular to nocturnal activity period. Both the head turning and wing steering response profiles were similar (Fig. 2).

The optomotor response to higher spatial frequencies decreased with light level. This suggests that the ability of M. sexta to resolve the motion of high spatial frequency decreases in dim light conditions. Also, the peak response to spatial frequency shifted to lower spatial frequency in moonlight. This could be due to the change in relative pupil size in dim light (Stöckl, 2016) and possible spatial summation in the visual system to improve the visual signal at dim light conditions (Warrant, 1999; Stöckl et al., 2016b). It has been suggested that the spatial summation is mediated by the lamina monopolar cells (LMCs) located in the lamina, the first visual processing area of the insect brain (Stöckl et al., 2016a). The hawkmoth LMCs have laterally extended arborizations likely to facilitate the spatial summation of visual signals from the neighboring LMCs to increase visual sensitivity at dim light conditions (Stöckl et al., 2016a). One possible consequence of this mechanism would be to decrease or eliminate the response to higher spatial frequencies.

Previous studies on the temporal tuning of LPTCs found that responses could be elicited from ∼1 to 7 Hz with peak response at 2 Hz (Theobald et al., 2009; Stöckl, 2016). In our moths' behavioral response, we observed a narrow temporal frequency tuning with a peak at 4.5 Hz. Although the optomotor response weakened as the light level reached moonlight, the peak response frequency remained the same throughout all light levels. The possible reasons for the observed differences in the peak temporal frequency between the experiments are discussed below.

A previous study using the closely related diurnal sphingid moth Macroglossum stellatarum showed a functional regionalization of the compound eye in processing rotational and translational motion stimuli (Kern and Varjú, 1998). This study also showed that the preferred temporal frequencies for rotational and translational visual motion stimuli were different, peaking at 2 and 4 Hz, respectively. While the frontal visual field was influential in shaping response to translational stimuli, the frontal and lateral visual fields of the compound eyes together were important in shaping the response to rotational stimuli (Kern and Varjú, 1998). Based on this study, we suggest that the behaviors associated with orienting to and tracking a moving flower during feeding rely on visual information detected in the frontal visual field, and thus would converge on the measured 2 Hz value (Kern and Varjú, 1998; Sponberg et al., 2015; Stöckl et al., 2016,b). In contrast, information detected with the lateral visual field may converge on the measured 4 Hz (Kern and Varjú, 1998) and be helpful in the side-to-side counterturning flight characteristic of odor plume tracking in M. sexta (Willis and Arbas, 1991).

In our study, the visual stimulus was projected onto a dome-shaped screen where the stimulus covered both frontal and lateral parts of the compound eye, simulating a rotational motion. As all previous studies presented the visual stimuli with a flat screen monitor positioned in front of the moth, it could have stimulated primarily the frontal portions of the compound eye (Theobald et al., 2009; Stöckl et al., 2016,b). One possible reason for the observed differences in the temporal frequency tuning between our results and the previous studies could be a result of the stimulus being perceived by the moths as rotational rather than translational. This needs to be tested.

For this study, we used tethered moth preparations to understand the effects of dim light on wide-field motion vision. This preparation enabled us to control the visual environment and quantify the effects of dim light on wide-field motion vision. However, the moth's freedom of movement was restricted and they were deprived of other important sensory inputs such as odor and airflow, as well as behavior-appropriate proprioceptive inputs that shape normal behavior. Recent studies show the importance of multisensory information in shaping flight control in the hawkmoth (Hinterwirth and Daniel, 2010; Dickerson et al., 2014; Windsor et al., 2014; Roth et al., 2016; Windsor and Taylor, 2017). In freely behaving moths, the complete suite of sensory inputs and the behavioral state might result in shorter latencies of response to wide-field motion (Olberg and Willis, 1990; Chiappe et al., 2010; Maimon et al., 2010).

The directionally selective wide-field cells in the lobula plate of insects have been shown to undergo motion adaptation, when exposed to high velocity even briefly, a duration as short as 20 ms (Maddess and Laughlin, 1985; Harris et al., 1999; Barnett et al., 2010; Nordström et al., 2011). This motion adaptation has been shown to reduce the LPTC contrast gain function and make them sensitive to changes in image velocity (Maddess and Laughlin, 1985; Harris et al., 1999; Barnett et al., 2010; Nordström et al., 2011). However, it has been shown that a interval of 2 s between the adaptation and test stimulus is sufficient to avoid the effects of motion adaptaion (Nordström et al., 2011). As each stimulus direction in this experiment was 3 s long at a constant velocity, this might have produced motion adaptation in the moths' motion vision pathway. Even though it is possible that motion adaptation could have occurred, our response speed measurements were made as the moths responded to the change in stimulus direction. Also, we had a 2 s inter-trial interval, which might have avoided any possible effects of motion adaption on the next trial. Taken together, these suggest that motion adaptation was probably not part of the responses we measured.

Dim light conditions have been shown to introduce a temporal lag in the ability of freely flying M. sexta to track a moving flower while hovering and nectar-feeding (Sponberg et al., 2015; Stöckl et al., 2017). However, for moths flying in dim light conditions, estimating visual flow field information is likely to be aided by a spatial summation strategy rather than a temporal summation strategy (Warrant, 1999). A temporal strategy is more suited for situations where the animal has to track slow-moving objects in dim light (Warrant, 1999; Sponberg et al., 2015). To estimate the effect of dim light on wide-field motion response in our experiments, we measured the time taken by the moth to follow the moving wide-field stimuli at a range of speeds and at different light levels (Figs 6 and 7). Our analysis revealed that at a given light level, the response time varied proportionally as a function of the speed of stimulus motion. To compare the response time across the light levels we used the estimated slope of the stimulus speed to response–time relationship. When compared across the light levels, there was no statistically significant change in response time. The slope value increased when the light level approached moonlight (0.01 cd m⁻²) and the response itself became weaker. We noticed that the relationship between the response speed and temporal frequency for stimuli with 5% contrast stimulus at moonlight level showed no linear relationship, and the slope value was significantly different from others for the response speed measurements and not for the onset latency measurements. At this light level, moths were selective in their response to temporal frequency and the response was weak at 5% contrast, hence fewer data points. Even though this is statistically significant (Fig. 7F), it may not be biologically significant. It remains to be verified how these wide-field motion responses would affect a plume-tracking moth in dim light condtions. In addition to these two measurements, we also looked at the time taken by the moths to respond to the change in stimulus motion direction (Fig. S5). The measured direction switch delay did not show any relationship to the temporal frequencies like the onset latency and response speed measurements (Fig. S5A–E). Also, when compared across the light levels and contrasts we did not observe any change in the estimated slope values (Fig. S5F). Based on our results, we suggest that visual acuity is affected as the light level decreases, indicating that the moth's ability to gather visual flow-field information could become noisier in dim light conditions. This decrease in spatial resolution could result in the slower behavioral responses observed in earlier studies.

In addition, freely flying plume-tracking moths will experience multiple directions of optic flow as they perform lateral movements and counterturns in three dimensions. During the lateral turn maneuvers, the compound eye on the leading side is likely to experience an expanding retinal image and the trailing side a contracting retinal image. As the animal changes direction, the compound eyes are then likely to experience an opposite effect. This could be interpreted by the visual system as a looming visual stimulus rather than, or in addition to, a lateral optic flow stimulus. Future experiments in free flight will determine how the moth's visual system processes complex optic flow fields in low light conditions.

We thank Eric Warrant and Anna Stöckl for their helpful comments and discussions on the project, and Jessica Fox and Fox laboratory members for their comments and suggestions on the manuscript.