Conditional Freezing, Flight and Darting?

Fear conditioning is one of the most frequently used laboratory procedures for modeling learning and memory generally, and anxiety disorders in particular. The conditional response (CR) used in the majority of fear conditioning studies in rodents is freezing. Recently, it has been reported that under certain conditions, running, jumping or darting replaces freezing as the dominant CR. These findings raise both a critical methodological problem and an important theoretical issue. If only freezing is measured but rodents express their learning with a different response, then significant instances of learning, memory, or fear may be missed. In terms of theory, whatever conditions lead to these different behaviors may be a key to how animals transition between different defensive responses and different emotional states. We replicated these past results but along with several novel control conditions. Contrary to the prior conclusions, running and darting were entirely a result of nonassociative processes and were actually suppressed by associative learning. Darting and flight were taken to be analogous to nonassociative startle or alpha responses that are potentiated by fear. On the other hand, freezing was the purest reflection of associative learning. We also uncovered a rule that describes when these movements replace freezing: When afraid, freeze until there is a sudden novel change in stimulation, then burst into vigorous flight attempts. This rule may also govern the change from fear to panic.


Introduction 38
Fear limits the behaviors available to an animal to its species-specific defense reactions 39 (SSDRs), thereby precluding more flexible voluntary behavior (Bolles, 1970). This characteristic 40 is one reason that conditions characterized by high fear levels such as anxiety disorders are so 41 maladaptive (Fanselow, 2018). It is also one reason that Pavlovian fear conditioning is so easy 42 to measure in the laboratory, one can simply measure innate defensive responses (i.e., SSDRs) 43 to diagnose fear and fear-related memory. This has made fear conditioning one of the major 44 rodent assays of learning, memory and anxiety disorders. Over the last four decades fear 45 conditioning studies have extensively used one of these defensive behaviors, freezing, more 46 it reduces the likelihood of detection and attack by a predator (Fanselow & Lester, 1988). 50

However, if rodents have multiple defensive responses, an important theoretical question is 51
what are the conditions that select between different SSDRs (Fanselow, 1997). An influential 52 model of SSDR selection applied to both humans and rodents is Predatory (or Threat) 53 Imminence Continuum theory, which states that qualitatively distinct defensive behaviors are 54 matched to the psychological distance from physical contact with a life-threatening situation 55 8 However, it is possible that during the initial test trials the response to the noise occurred via 162 second-order conditioning as the noise was paired with the previously reinforced tone. This 163 seems unlikely because most darts were seen at the beginning of testing and decreased over 164 the session. A second-order conditioning interpretation suggests the opposite pattern. 165 Nonetheless, in a second experiment, we included classic controls to directly test for the 166 phenomenon of pseudo-conditioning (Table 2). Pseudo-conditioning is a form of sensitization 167 whereby mere exposure to the US changes behavior to the stimulus used as a CS (Underwood, 168 1966), and this appears to be what was observed in Experiment 1 (Stimulus Change Group; 169 in the prior study without any auditory stimuli (no CS). A third was merely exposed to the 171 chamber. The final group was a conditioning group that received noise-shock pairings. All 172 groups received tests with the 10 sec noise, except for one of the pseudoconditioning groups 173 that was tested with the tone. 174 Figures 4 and 5 summarize the test results from Experiment 2 (see Fig. S4 for trial-by-trial data). 175 As would be expected for a CR, freezing to the noise was greatest in the mice that received 176 noise-shock pairings [F(3,28) = 11.76, p<.001]. Significant associative learning was indicated 177 by more noise-elicited freezing in the paired group than the shock-only trained group tested with 178 the noise. Interestingly, the No Shock group that was tested with the noise gradually increased 179 freezing over the course of noise testing (Fig. S4) suggesting that the 75dB noise itself was 180 aversive to the mice and could support some conditioning of freezing (i.e., it was a weak US). 181 The test session data were very different for activity bursts (Figs. 4 and 5). The greatest PAR 182 occurred in the pseudoconditioned control (shock only during training) that was tested with the 183 novel noise [F(3,28) = 20.085, p<.001]. The pseudoconditioned control tested with the novel 184 noise showed the most darting behavior. Furthermore, these results are supported by a direct 185 analysis of velocity data during the 10s CS period at test (Fig. 5) In a third experiment, we included a control group in which the shock and noise were explicitly 204 unpaired to again test for the phenomenon of pseudo-conditioning but in a situation where 205 exposure to the CS is equated during training (Table 3). One group was again a conditioning 206 group that received noise-shock pairings, and one group was again a pseudoconditioned group 207 that only received shocks without any CS. One group received equal numbers of noise and 208 shock presentations but in an explicitly unpaired manner. An additional control group received 209 presentations of only the white noise CS to examine whether or not the CS alone was able to 210 support conditioning and/or activity bursts. 211 Acquisition and test results are summarized in Figures 6 and 7 That pairing noise and shock altered the timing of the activity bursts is an interesting fact worth 240 considering and suggests that pairing noise and shock may have primarily resulted in a 241 conditioned freezing response which in fact competes with/reduces any initial non-associative 242 activity/bursting to the white noise. Taken together, this and the prior experiment using control 243 groups to assess pseudoconditioning reveal that a large portion, if not all, of the noise-elicited 244 activity bursts observed are due to non-associative processes which result in an increase in 245 darting behavior to the noise following shock exposure, regardless of any direct training history 246 of the noise with shock. There does appear to be evidence that pairing noise with shock may 247 further increase or alter the timing of this behavior, but by no means is pairing noise with shock 248 necessary to produce these activity bursts. 249 250

Experiment 4 251
The experiments thus far have suggested that much of the white-noise-elicited activity bursting 252 is a non-associative process. We have also shown that novelty of the CS at test may increase 253 this noise-elicited activity (Figs. 3 & 4). In a final, fourth experiment, we explicitly tested whether 254 habituation to the white noise stimulus prior to noise-shock training would be able to reduce 255 noise-elicited activity bursts. If increased levels of novelty of the CS are driving noise-elicited 256 activity bursts, then prior habituation should reduce the levels of darting to the noise CS. In this 257 experiment, we had four groups which differed in whether they received an additional two days 12 of habituation to the white noise stimulus (5 noise presentations each day) and whether they 259 received noise-shock pairings during training or just shock only (Table 4). One comparison of 260 particular interest was between the habituated or non-habituated Shock Only groups as these 261 groups would directly compare whether prior experience with the CS would decrease darting at 262 test compared to a group for which the CS was completely novel. 263 Figure 8 shows the results of Experiment 4 during testing (see Figure S5 for trial-by-trial results 264 for freezing, PAR, and darting across habituation, training, and testing). During the two days of 265 habituation, interestingly, we found that within groups which received habituation, a low level of showing that this white noise stimulus alone can act as a US. It is interesting that darting 270 occurred to the white noise at the start of habituation when the CS was very novel, and at the 271 end of habituation once the white noise alone was able to support some level of fear. 272 Comparing the two Shock Only groups during test, the noise disrupted freezing more than tone. 273 In this regard noise seems to act like a weak shock US (Fanselow, 1982). Like shock it disrupts 274 freezing ( Fig S5) and like shock it supports conditioning of freezing (Fig 6).  We again saw that freezing to the white noise initially increased during acquisition, but as the 280 darting response begins to become more apparent, freezing decreases to medium levels. At such that animals who received white noise paired with shock froze more than animals who only 283 13 received shock during acquisition, again indicative that noise-elicited freezing is a conditional 284 behavior that results from associative learning. For darting behavior, we found a Habituation X 285 Pairing interaction [F(1,28)=4.939, p=.035] such that pairing white noise with shock increased 286 darting within habituated animals (p=.033), and that habituation reduced darting within animals 287 who only received shock during training (p=.045). These results reveal multiple points of 288 interest. First, and as shown in prior experiments, the white noise acts as a US on its own and 289 need not be paired with shock to produce darting at test. Merely experiencing the shock is 290 enough to produce darting to the white noise at test (pseudoconditioning due to sensitization). 291 Furthermore, prior experience with the white noise, through habituation, actually reduced this 292 darting at test. Additionally, in this experiment, we do again show evidence that pairing white 293 noise with shock can further increase darting behavior at test, at least within animals who have 294 already experienced the noise during habituation. Again, as with Experiment 3 (Fig. 7) the 295 timing of the darting response in Paired groups is fundamentally altered compared to Shock 296 Only groups (Fig. 8). The magnitude/frequency of the initial activity burst to the noise appears 297 to be reduced in the Paired groups, and increased levels of activity bursts during the latter 298 portion of the CS account for any differences/increases in overall numbers of darts. noise period and then quickly decreased to more stable levels. As seen in the experiments 304 above, again, this initial peak in velocity was most apparent in the Shock Only groups, with the 305 Paired groups showing an initially smaller peak in velocity. Post-hoc analyses revealed that the 306 Shock Only groups had significantly higher velocity during the first three bins of the noise than 307 the Paired groups (p's=.02, .03, .005 respectively). Post-hoc analysis on the Pairing X 308 Habituation interaction reveal that within the non-habituated groups, pairing noise and shock 309 significantly reduced the velocity throughout test trials (p<.001). Additionally, within Shock Only Prior work reported that contact/pain-related stimuli (e.g., shock) disrupt freezing and provoke 317 panic-like circa-strike defensive behaviors (Fanselow, 1982). The current results suggest a 318 modification of the rules governing a transition between these behavioral states. The rule is that 319 when you are in the post-encounter mode (fear) a sudden change in stimulation, particularly the 320 onset of an intense novel stimulus, can cause an immediate transition to the circa-strike mode 321 (panic). Indeed, the vast majority of the activity bursts/darting behavior occurred at the onset of 322 the stimulus (Figs. 3, 5, 7, 8). The effectiveness of this transition depends on the qualities of the 323 stimulus. Stronger shocks cause a greater disruption of freezing and a longer activity burst, yet 324 the same stronger shocks simultaneously condition more freezing to the prevailing cues 325 (Fanselow, 1982). The current data call for an expansion of this rule to nonnociceptive stimuli. 326 While both tone and noise disrupted ongoing freezing, the noise did so for longer than the tone 327 Our interpretation that noise unconditionally elicits a ballistic activity burst bears some 362 relationship to the unconditional acoustic startle response. Loud noises will elicit an 363 unconditional startle response that wanes with repeated presentations of that noise (i.e., 364 habituation; e.g., Davis, 1980;Hoffman & Fleshler, 1963;Leaton, 1976). While our 75 dB noise 365 stimulus is less intense than the 98-120 dB noise used in typical acoustic startle studies, we are 366 observing an unconditional noise elicited response that also decreases with habituation 367 (Experiment 4). Furthermore, our data and those of Totty et al. (2021) indicate that these 368 responses require a fearful context in order to occur. Fear is well known to potentiate startle 369 responses (Brown, et al., 1951;Davis, 1989). Perhaps the low intensity noise is below 370 threshold to elicit a startle response on its own, but a fearful context potentiates this response 371 and brings it above threshold. Additionally, there appears to be considerable overlap in the 372 neuroanatomy that supports this circa-strike behavior and fear potentiated startle. Totty et al.  It is of note that the relationship between startle (circa-strike defense) and freezing (post-381 encounter defense) was described by Fanselow & Lester (1988) when accounting for how rats 382 rapidly transitioned between these behaviors when a detected predator launches into attack. "It 383 is as if the freezing animal is tensed up and ready to explode into action if the freezing response 384 fails it. This explosive response probably has been studied in the laboratory for over 30 years 385 under the rubric of potentiated startle…It seems that the releasing stimulus for this explosive 386 motor burst is a sudden change in the stimulus context of an already freezing rat (Fanselow & 387 Lester, 1988 behavior, which is something required in order to conclude that a response is conditional 390 (Rescorla, 1967). Both of these research groups concluded from their single group experiments 391 that flight/darting was a CR because the behavior increased with successive shocks during the 392 shock phase and decreased with shock omission during the test phase, likening these 393 behavioral changes to acquisition and extinction. While acquisition and extinction are 394 characteristics of a CR, learning theorists have never taken these as diagnostic of a CR. For 395 example, increases in responding with successive shocks could arise via sensitization and 396 decreases in responding when shocks are omitted could arise from habituation. Indeed, that is 397 exactly what we believe caused these behavioral changes that we also observed in our study. noise-shock paired rats showed more noise elicited activity burst behavior than rats that had 406 unpaired noise and shock requires additional comment. Since both unpaired and paired rats 407 were exposed to noise during acquisition those exposures could lead to habituation of the 408 unconditional response to the noise. However, it would be expected that habituation would be 409 greater in the unpaired group because pairing a stimulus (noise in this case) with another 410 stimulus (shock in this case) is known to reduce the magnitude of habituation (Pfautz et al., 411 1978). This reduction in habituation is observed even if the second stimulus is not an 412 unconditional stimulus (Pfautz et al., 1978). Additionally, pairing a habituated stimulus with a US 413 can also cause a return of the habituated alpha response and this loss of habituation is not 414 observed when the two stimuli are not paired (Holland, 1977). Thus, the difference between the the best thing for a small mammal like a rat or a mouse to do when a predator is detected and 435 will only be replaced if there is a change consistent with contact (Fanselow & Lester, 1988). 436 Rodents choose locations in which to freeze such as corners or objects (thigmotaxis) (Grossen 437 & Kelley, 1972). The current data show that the freezing rodent also prepares to react to 438 sudden stimulus change. There is nothing passive about it. 439

Subjects 441
Subjects for all experiments included 120 C57BL/6NHsd mice (Experiment 1, n=24; Experiment 442 2, n=32; Experiment 3, n=32; Experiment 4, n=32), aged 9-11 weeks of age and purchased 443 from Envigo. This C57BL/6NHsd strain was chosen to match that of Experiment 1 was conducted as delineated in Table 1 (see Fig. 1 for a schematic representation 475 of the serial conditioned stimulus and the design for training and testing for Experiment 1). The 476 Replication group was trained on each of the two days with 5 presentations of a 10 second tone 477 immediately followed by a 10 second noise, which was immediately followed by a 1 second 478 shock. On Day 3 it was then tested with 16 presentations of a 10 second tone immediately 479 followed by a 10 second noise. These parameters were chosen to match those of Fadok et al 480 (2017) except that we did not include a session of unreinforced CS preexposure prior to 481 conditioning as such treatment is known to reduce conditioned behavior (Lubow & Moore, 1959;482 we did add such a treatment to Experiment 4 as an experimental factor). The CS Duration group 483 was trained on each of the two days with 5 presentations of a 10 second noise, which was 484 immediately followed by a 1 second shock. It was tested with 16 presentations of the 10 second 485 noise. The Stimulus Change group was trained on each of the two days with 5 presentations of 486 a 20 second tone immediately followed by a 1 second shock. It was tested with 16 487 presentations of a 10 second tone immediately followed by a 10 second noise (i.e., the 488 22 compound used in the replication group). Two mice were excluded from this study due to 489 experimenter error, one female in the Replication group and one female in the Stimulus Change 490

group. 491
Experiment 2 was conducted as delineated in Table 2. The Pseudoconditioned Noise and 492 Pseudoconditioned Tone groups were trained on each of the two days with 5 presentations of a 493 1-sec shock without any sound using the same schedule for shocks as Experiment 1. The No 494 Shock Control was merely allowed to explore the context for the same length of time as the 495 other groups without receiving any shock or auditory stimuli throughout the two days of 496 acquisition. The final Noise-Shock Conditioning group was trained on each of the two days with 497 5 presentations of a 10-sec noise, which was immediately followed by a 1-sec shock. As 498 Experiment 1 revealed that similar behavior was observed in groups which received compound 499 stimulus-shock pairings or just noise-shock pairings, we utilized simple noise-shock pairings in 500 this and some of the following experiments to more specifically assess the associative nature of 501 any white noise-driven behavior. All groups received tests with 16 presentations of the 10-sec 502 noise in extinction, except for one of the pseudoconditioning groups that was tested with the 10-503 sec tone. 504 Experiment 3 was conducted as delineated in Table 3. The Paired Noise-Shock (Conditioning) 505 group was trained on each of the two days with 5 presentations of a 10 second noise, which 506 was immediately followed by a 1 second shock. The Unpaired Noise/Shock group was 507 presented with the same number and length of noise and shocks, but they were explicitly 508 unpaired in time. The Noise-CS Only group received 5 presentations of a 10 second noise 509 without receiving any shocks on each of the two days. The Shock Only (Pseudoconditioning) 510 group received 5 presentations of a 1 second shock on each of the two days. As the main 511 behavioral responses and differences between groups occurred primarily in the first few trials of 512 the previous experiments, and in order to more readily complete all of the testing within one 513 day's light cycle, for this and the following experiments we reduced the number of test trials 514 presented to the animals. Thus, at test for this experiment, all groups received two 515 presentations of a 10 second noise. 516 Experiment 4 was conducted as delineated in Table 4. Prior to training with shock, all groups 517 underwent 2 days of additional training with either habituation to the white noise or merely 518 exposure to the context. The habituated groups, Habituation/Shock Only (H-Shock) and 519 Habituation/Noise-Shock Pairing (H-Paired), were trained on each of the two days with 5 520 presentations of a 10-second noise, while the two non-habituated groups, Context 521 Exposure/Shock Only (C-Shock) and Context Exposure/Noise-Shock Pairing (C-Paired) 522 received only equivalent exposure to the context. The following two days, and as in the 523 Experiments above, all groups received 10 footshocks. The Paired groups (H-Paired and C-524 Paired) were trained on each of the two days with 5 presentations of a 10-second noise, 525 followed immediately by a 1-second footshock. The Shock Only groups (H-Shock and C-Shock) 526 were trained on each of the two days with only 5 presentations of a 1-second footshock. At test, 527 all groups received 3 presentations of the 10-second noise. 528

Data and Statistics and Analysis 529
Freezing behavior for Experiments 1-3 was scored using the near-infrared VideoFreeze scoring 530 system. Freezing is a complete lack of movement, except for respiration (Fanselow, 1980). 531 VideoFreeze allows for the recording of real-time video at 30 frames per second. With this 532 program, adjacent frames are compared to provide the grayscale change for each pixel, and the 533 amount of pixel change across each frame is measured to produce an activity score. We have 534 set a threshold level of activity for freezing based on careful matching to hand-scoring from 535 trained observers (Anagnostaras et al., 2010). The animal is scored as freezing if they fall 536 below this threshold for at least a 1-sec bout of freezing. 537 For Experiment 4, due to a technical error, videos for the first 4 days of the experiment could not 538 be accurately assessed for freezing behavior using VideoFreeze. Therefore, we alternatively 539 measured and scored freezing behavior using EthoVision. Briefly, videos were converted to 540 MPEG, as described above, and analyzed using the Activity Analysis feature of Ethovision. 541 Thresholds for freezing were again determined to match hand-scoring from trained observers. 542 Two different measures of flight were used. We scored bursts of locomotion and jumping with a 543 Peak Activity Ratio (PAR); Fanselow et al., 2019) and the number of darts (Gruene et al., 2015). 544 To determine PAR, we took the greatest between frame activity score during a period of interest 545 (e.g., the first 10 s of CS presentation = During) and calculated a ratio of that level of activity to a 546 similar score derived from a preceding control period of equal duration (e.g., 10 s before 547 presentation of the tone = PreStim) of the form During/(During + PreStim). With this measure, a 548 0.5 indicates that during the time of interest there was no instance of activity greater than that 549 observed during the control period (PreStim). PARs approaching 1.0 indicate an instance of 550 behavior that far exceeded baseline responding. This measure reflects the maximum movement 551 the animal made during the period of interest. 552 Darting was assessed as in Gruene et al. (2015). Video files from VideoFreeze were extracted 553 in Windows Media Video format (.wmv) and then converted to MPEG-2 files using Any Video 554 Converter (AnvSoft, 2018). These converted files were then analyzed to determine animal 555 velocity across the session using EthoVision software (Noldus), using a center-point tracking 556 with a velocity sampling rate of 3.75 Hz. This velocity data was exported, organized, and 557 imported to R (R Core Team, 2018). Using a custom R code (available as source code 1), darts 558 were detected in the trace with a minimum velocity of 22.9 cm/s and a minimum interpeak 559 interval of 0.8 s. The 22.9 cm/s threshold was determined by finding the 99.5th percentile of all 560 baseline velocity data analyzed, prior to any stimuli or shock, and this threshold was validated to 561 match with manual scoring of darts, such that all movements at that rate or higher were 562 consistently scored as darts. See Figure S1 for representative traces of velocity across Day 1 563 of acquisition for a mouse in the Replication group of Experiment 1. The PAR measure reflects 564 the maximum amplitude of movement, while the dart measure reflects the frequency of 565 individual rapid movements.      for Experiment 1. Data is averaged across all animals per group and binned into ~.5s bins 604 (0.533s) and presented as means plus/minus standard error (Mean ±SE). These within-subject 605 error bars are corrected for between-subject variability using methods as described in Rouder 606 and Morey (2005). During this test, the Replication Group and the Stimulus Change Group 607 received the serial conditioned stimulus (SCS) in which a 10s tone was followed by a 10s noise. 608 The CS Duration group was only tested with a 10s noise. 609 for Experiment 2. Data is averaged across all animals per group and binned into ~.5s bins 617 (0.533s) and presented as means plus/minus standard error (Mean ±SE). These within-subject 618 error bars are corrected for between-subject variability using methods as described in Rouder 619 and Morey (2005). During this test, the No Shock-Noise Test, Shock Only-Noise Test, and 620 Noise-Shock Noise Test groups were tested with a 10s noise. The Shock Only-Tone Test group 621 was tested with a 10s tone. 622  that the occurrence of the stimuli at test disrupt freezing to the context and that the noise 665 disrupts freezing to a greater extent than the tone. 666