Hoerl & McCormack (H&M) propose that two systems guide behavior in time. Both human and nonhuman animals possess the “temporal updating” system; it forms an internal model of the world based on how events change across time. Through experience, the model becomes entrained to temporal patterns in the environment. When a sequence of events begins, the model is continuously updated in accordance with prior experience. If these updates coincide with changes in the environment, then the animal will act appropriately in time. Importantly, the dimension of time is not explicitly represented within the temporal updating system. Changes in the model simply determine the animal's current expectations, similar to how gravitational changes from the moon determine current ocean tides (cf. Killeen & Fetterman Reference Killeen and Fetterman1988). However, the “temporal reasoning” system is held to be uniquely possessed by humans; here, time is represented continuously and the temporal locations of events are explicitly tracked.
Overall, the proposal brings a useful perspective to key findings related to whether animals are “stuck in time” (Clayton & Dickinson Reference Clayton and Dickinson1998; Roberts Reference Roberts2002). Our primary issue with this hypothesis concerns claims about how humans and nonhuman animals engage in sequential learning. According to the authors, humans learn sequences by reasoning about the temporal relationships that separate events. However, nonhuman animals cannot represent the “temporal addresses” of prior events. Therefore, they must engage in sequential learning by being exposed to the events within a sequence in their “correct and full order,” allowing the model to entrain to the sequence. The authors propose that this is a “signature limit” of the temporal updating system. Nonetheless, this claim does not comport with prior work.
For example, research suggests that nonhuman animals generate “temporal maps” of sequential events (Balsam & Gallistel Reference Balsam and Gallistel2009); that is, they represent time as a continuum and are able to arrange the temporal locations of events in a sequence along this array. This work parallels the less controversial claim that animals generate “spatial maps” of the environment that they use to remember the locations of objects in space.
Temporal map research is extensive and longstanding (Honig Reference Honig, Spear and Miller1981; Matzel et al. Reference Matzel, Held and Miller1988), but a simple experimental example runs as follows. During Phase 1, animals are placed into a conditioning chamber and repeatedly presented with two cues that are separated by a certain time interval (Cue A→Cue B). On a subsequent training day (Phase 2), animals return to the chamber and are presented with a reward that is followed soon after by the second cue from Phase 1 (Reward→Cue B). Importantly, the time interval separating reward and Cue B is arranged so that, in the context of Phase 1, reward should have occurred at a certain time between the cues (Cue A→Reward→Cue B). However, the animals never explicitly experience the full sequence in its appropriate order; they only experience misordered fragments of the full, “implied” sequence. Yet, animals appear to integrate the two learning episodes and behave according to the implied sequence. When given Cue A, they expect reward, despite never receiving reward after its presentation. Conversely, when presented with Cue B, they do not expect reward.
Various permutations of this basic design have extended this effect (Molet & Miller Reference Molet and Miller2014), which we cannot detail for brevity. However, these experiments strongly suggest that animals not only order the shuffled fragments of an implied sequence, they also represent the continuous intervals that separate each event it contains (Molet et al. Reference Molet, Miguez, Cham and Miller2012; Taylor et al. Reference Taylor, Joseph, Zhao and Balsam2014). In other words, animals represent the temporal locations of items within a continuous temporal map, much like they track the spatial locations of items within a spatial map. These findings are difficult to reconcile with the authors’ claim that nonhuman animals should only be able to learn a sequence by explicitly experiencing every event within the pattern in its appropriate order.
A related line of work suggests that nonhuman animals understand the concept of ordinal position (Orlov et al. Reference Orlov, Yakovlev, Hochstein and Zohary2000; Orlov et al. Reference Orlov, Yakovlev, Amit, Hochstein and Zohary2002). In these experiments, animals learn the order of different lists of items (e.g., List 1: A1, A2, A3 / List 2: B1, B2, B3). Then, animals are simultaneously presented with items from one of the lists and are required to tap each in its appropriate order (e.g., first A1, then A2, then A3). Importantly, a distractor is also presented, composed of an item in one of the other lists (e.g., B2). Animals frequently tap the distractor. Importantly, they do so systematically, usually corresponding to the distractor's ordinal position within its own list (e.g., responding: A1→B2→A3, rather than the correct pattern: A1→A2→A3). Furthermore, if animals later learn novel lists containing items from initial training, they learn faster when each item maintains its previous ordinal position (e.g., new list: A1→B2→C3 vs. C3→A1→B2; Chen et al. Reference Chen, Swartz and Terrace1997). Carefully controlled experiments have ruled out accounts of performance being solely based on rote long-term memory, working memory, and simple associative-linking processes. Again, these findings are difficult to reconcile with the claim that animals learn sequences by repeatedly updating an internal model as items are experienced across time.
Given the emphasis on sequential learning processes as a primary difference between the two systems, we were surprised that H&M did not address these findings. To be clear, we are not arguing that animals possess the temporal reasoning system. Associative explanations may adequately explain the above findings, albeit ones that cede a basic time representation to nonhuman animals (Arcediano & Miller Reference Arcediano and Miller2002). However, representing temporal information does not necessarily imply that nonhuman animals are capable of mental time travel (for a discussion of this topic in relation to temporal maps, see Arcediano & Miller Reference Arcediano and Miller2002) or even retain a timeline of life experiences. For example, after integrating two learning episodes during a temporal map experiment, does the animal still remember that the two learning episodes were experienced apart from one another? Or have they formed an immutable model of the implied sequence (with a time dimension) effectively “losing” knowledge of the component learning episodes? The work reviewed here does not speak to these questions. All of the reviewed work requires either a qualification or clarification of the authors’ hypothesis.
You have
Access
Hoerl & McCormack (H&M) propose that two systems guide behavior in time. Both human and nonhuman animals possess the “temporal updating” system; it forms an internal model of the world based on how events change across time. Through experience, the model becomes entrained to temporal patterns in the environment. When a sequence of events begins, the model is continuously updated in accordance with prior experience. If these updates coincide with changes in the environment, then the animal will act appropriately in time. Importantly, the dimension of time is not explicitly represented within the temporal updating system. Changes in the model simply determine the animal's current expectations, similar to how gravitational changes from the moon determine current ocean tides (cf. Killeen & Fetterman Reference Killeen and Fetterman1988). However, the “temporal reasoning” system is held to be uniquely possessed by humans; here, time is represented continuously and the temporal locations of events are explicitly tracked.
Overall, the proposal brings a useful perspective to key findings related to whether animals are “stuck in time” (Clayton & Dickinson Reference Clayton and Dickinson1998; Roberts Reference Roberts2002). Our primary issue with this hypothesis concerns claims about how humans and nonhuman animals engage in sequential learning. According to the authors, humans learn sequences by reasoning about the temporal relationships that separate events. However, nonhuman animals cannot represent the “temporal addresses” of prior events. Therefore, they must engage in sequential learning by being exposed to the events within a sequence in their “correct and full order,” allowing the model to entrain to the sequence. The authors propose that this is a “signature limit” of the temporal updating system. Nonetheless, this claim does not comport with prior work.
For example, research suggests that nonhuman animals generate “temporal maps” of sequential events (Balsam & Gallistel Reference Balsam and Gallistel2009); that is, they represent time as a continuum and are able to arrange the temporal locations of events in a sequence along this array. This work parallels the less controversial claim that animals generate “spatial maps” of the environment that they use to remember the locations of objects in space.
Temporal map research is extensive and longstanding (Honig Reference Honig, Spear and Miller1981; Matzel et al. Reference Matzel, Held and Miller1988), but a simple experimental example runs as follows. During Phase 1, animals are placed into a conditioning chamber and repeatedly presented with two cues that are separated by a certain time interval (Cue A→Cue B). On a subsequent training day (Phase 2), animals return to the chamber and are presented with a reward that is followed soon after by the second cue from Phase 1 (Reward→Cue B). Importantly, the time interval separating reward and Cue B is arranged so that, in the context of Phase 1, reward should have occurred at a certain time between the cues (Cue A→Reward→Cue B). However, the animals never explicitly experience the full sequence in its appropriate order; they only experience misordered fragments of the full, “implied” sequence. Yet, animals appear to integrate the two learning episodes and behave according to the implied sequence. When given Cue A, they expect reward, despite never receiving reward after its presentation. Conversely, when presented with Cue B, they do not expect reward.
Various permutations of this basic design have extended this effect (Molet & Miller Reference Molet and Miller2014), which we cannot detail for brevity. However, these experiments strongly suggest that animals not only order the shuffled fragments of an implied sequence, they also represent the continuous intervals that separate each event it contains (Molet et al. Reference Molet, Miguez, Cham and Miller2012; Taylor et al. Reference Taylor, Joseph, Zhao and Balsam2014). In other words, animals represent the temporal locations of items within a continuous temporal map, much like they track the spatial locations of items within a spatial map. These findings are difficult to reconcile with the authors’ claim that nonhuman animals should only be able to learn a sequence by explicitly experiencing every event within the pattern in its appropriate order.
A related line of work suggests that nonhuman animals understand the concept of ordinal position (Orlov et al. Reference Orlov, Yakovlev, Hochstein and Zohary2000; Orlov et al. Reference Orlov, Yakovlev, Amit, Hochstein and Zohary2002). In these experiments, animals learn the order of different lists of items (e.g., List 1: A1, A2, A3 / List 2: B1, B2, B3). Then, animals are simultaneously presented with items from one of the lists and are required to tap each in its appropriate order (e.g., first A1, then A2, then A3). Importantly, a distractor is also presented, composed of an item in one of the other lists (e.g., B2). Animals frequently tap the distractor. Importantly, they do so systematically, usually corresponding to the distractor's ordinal position within its own list (e.g., responding: A1→B2→A3, rather than the correct pattern: A1→A2→A3). Furthermore, if animals later learn novel lists containing items from initial training, they learn faster when each item maintains its previous ordinal position (e.g., new list: A1→B2→C3 vs. C3→A1→B2; Chen et al. Reference Chen, Swartz and Terrace1997). Carefully controlled experiments have ruled out accounts of performance being solely based on rote long-term memory, working memory, and simple associative-linking processes. Again, these findings are difficult to reconcile with the claim that animals learn sequences by repeatedly updating an internal model as items are experienced across time.
Given the emphasis on sequential learning processes as a primary difference between the two systems, we were surprised that H&M did not address these findings. To be clear, we are not arguing that animals possess the temporal reasoning system. Associative explanations may adequately explain the above findings, albeit ones that cede a basic time representation to nonhuman animals (Arcediano & Miller Reference Arcediano and Miller2002). However, representing temporal information does not necessarily imply that nonhuman animals are capable of mental time travel (for a discussion of this topic in relation to temporal maps, see Arcediano & Miller Reference Arcediano and Miller2002) or even retain a timeline of life experiences. For example, after integrating two learning episodes during a temporal map experiment, does the animal still remember that the two learning episodes were experienced apart from one another? Or have they formed an immutable model of the implied sequence (with a time dimension) effectively “losing” knowledge of the component learning episodes? The work reviewed here does not speak to these questions. All of the reviewed work requires either a qualification or clarification of the authors’ hypothesis.