First, what is being updated during temporal updating?
The authors elaborate on specific constraints of the temporal updating process, but they are less clear on the mental representation that is being updated, which they refer to as a “world model.” The world model they refer to bears resemblance to “perceptual mental models” described in research on event segmentation, visual perception, and mental simulation (Chua et al. Reference Chua, Boland and Nisbett2005; Churchland et al. Reference Churchland, Ramachandran, Sejnowski, Koch and Davis1994). The distinction offered by H&M, that perceptual models and event models may be fundamentally distinct in both evolutionary and developmental terms, could help frame current theories of event cognition so long as the factors that distinguish the two are clearly delineated.
Our own recent work (Kelly Reference Kelly2018; Khemlani et al. Reference Khemlani, Mackiewicz, Bucciarelli and Johnson-Laird2013; Reference Khemlani, Harrison and Trafton2015) can help distinguish perceptual models – which the temporal updating system produces – from event models, which are constructed during temporal reasoning. Some fundamental differences between perceptual models and event models are provided in Table 1. The table shows that both perceptual and event models are iconic, discrete simulations that represent a possible set of relations between entities. But while perceptual models come from using perceptual information to update a model of a reasoner's surroundings, event models can represent situations apart from the reasoner's ongoing experience. They can come from discourse concerning real or hypothetical scenarios that are spatiotemporally displaced; episodic memory of events in the past; and imagination about events in the future. Unlike perceptual models, event models can concern multiple situations. Consider the following description of a set of events: “The commute happened before the staff meeting. The commute happened before the conference call.” The description is consistent with at least two temporal possibilities: one in which the meeting happened before the call, and one in which the call happened before the meeting. Those who fail to enumerate the different possibilities will fail to grasp the ambiguity of the description (Kelly & Khemlani Reference Kelly, Khemlani, Goel, Seifert and Freksa2019). Event models permit reasoners to enumerate multiple possibilities.
Table 1. Conceptual and computational differences between the perceptual and event models
The table lists additional ways in which we believe perceptual models differ from event models. The differences are anticipated in part by H&M, who argue that “the temporal updating system … deals with changing input by changing representations, rather than by representing change” (sect. 1.1, para. 1). If H&M are right that temporal updating is a highly constrained cognitive process, then the representations it updates should be constrained in systematic ways that yield testable empirical predictions.
Second, is a dual-systems framework appropriate?
When theorists invoke a dual-systems account of reasoning, one fundamental assumption is that the two systems compute the same function in two different ways: an initial, rapid system computes a heuristic response based on one or more cues, and a slower, deliberative system processes the same information in a more elaborate manner (Stanovich & West Reference Stanovich and West2000). The two systems rely on different algorithms to carry out the same cognitive task. But when H&M distinguish updating from reasoning, the goals of the two systems they posit differ: People update their perceptual models to maintain an accurate simulation of reality. In contrast, a person may engage in temporal reasoning to achieve many different goals, for example, planning for the future, reinterpreting the past, comprehending discourse, and understanding the sequence of a film. Because temporal updating and temporal reasoning are used for different purposes, invoking the dual-systems framework may be inappropriate.
Indeed, it is not clear to us why updating and reasoning are the only processes relevant to temporal cognition. Some tasks that require the representation of time do not require reasoning at all. Consider the task of event recall (Wang & Gennari Reference Wang and Gennari2019). The task requires an individual to recall events that comprise some temporal interval. For example, you might summarize your previous day as follows: “I had breakfast, worked on a project, taught a class, had a meeting, then had lunch with a friend….” The task requires individuals to remember and then to represent multiple events along a mental timeline. It does not concern temporal updating and it does not require reasoning, either, because responders need not infer any novel temporal relations while recalling events in memory. The act of remembering a temporal sequence seems fundamental to temporal thinking, but the dual-systems framework that H&M espouse has no place for it.
Hence, H&M must explain whether their account allows for cognitive processes that result in mental representations of temporally ordered events, even those that do not demand explicit temporal reasoning. The “intermediate developmental stage” (sect. 3, para. 1) to which they refer presents a broad challenge to the dual-systems framework. Children may struggle to retrieve temporal sequences, not because they revert to updating, but rather because of episodic memory retrieval failures (Prabhakar & Ghetti Reference Prabhakar and Ghetti2019). H&M should enumerate the specific pattern of errors predicted by reverting to the updating system. Perhaps a more accommodative framework, one that retains the division outlined by H&M, should specify the different processes relevant to temporal cognition (e.g., updating, recall, reasoning) as well as the various representational and computational constraints of each process (cf. Khemlani et al. Reference Khemlani, Harrison and Trafton2015).
In sum, H&M's distinction between temporal updating and reasoning is useful, so much so that it is worth refining, clarifying, and extending to address the two issues highlighted above.
You have
Access
Hoerl & McCormack (H&M) distinguish between temporal updating and temporal reasoning as separate mental processes. The distinction is sensible and useful, and it helps synthesize many extant results in temporal cognition. Nevertheless, the authors’ framework prompts two issues worth clarifying:
First, what is being updated during temporal updating?
The authors elaborate on specific constraints of the temporal updating process, but they are less clear on the mental representation that is being updated, which they refer to as a “world model.” The world model they refer to bears resemblance to “perceptual mental models” described in research on event segmentation, visual perception, and mental simulation (Chua et al. Reference Chua, Boland and Nisbett2005; Churchland et al. Reference Churchland, Ramachandran, Sejnowski, Koch and Davis1994). The distinction offered by H&M, that perceptual models and event models may be fundamentally distinct in both evolutionary and developmental terms, could help frame current theories of event cognition so long as the factors that distinguish the two are clearly delineated.
Our own recent work (Kelly Reference Kelly2018; Khemlani et al. Reference Khemlani, Mackiewicz, Bucciarelli and Johnson-Laird2013; Reference Khemlani, Harrison and Trafton2015) can help distinguish perceptual models – which the temporal updating system produces – from event models, which are constructed during temporal reasoning. Some fundamental differences between perceptual models and event models are provided in Table 1. The table shows that both perceptual and event models are iconic, discrete simulations that represent a possible set of relations between entities. But while perceptual models come from using perceptual information to update a model of a reasoner's surroundings, event models can represent situations apart from the reasoner's ongoing experience. They can come from discourse concerning real or hypothetical scenarios that are spatiotemporally displaced; episodic memory of events in the past; and imagination about events in the future. Unlike perceptual models, event models can concern multiple situations. Consider the following description of a set of events: “The commute happened before the staff meeting. The commute happened before the conference call.” The description is consistent with at least two temporal possibilities: one in which the meeting happened before the call, and one in which the call happened before the meeting. Those who fail to enumerate the different possibilities will fail to grasp the ambiguity of the description (Kelly & Khemlani Reference Kelly, Khemlani, Goel, Seifert and Freksa2019). Event models permit reasoners to enumerate multiple possibilities.
Table 1. Conceptual and computational differences between the perceptual and event models
The table lists additional ways in which we believe perceptual models differ from event models. The differences are anticipated in part by H&M, who argue that “the temporal updating system … deals with changing input by changing representations, rather than by representing change” (sect. 1.1, para. 1). If H&M are right that temporal updating is a highly constrained cognitive process, then the representations it updates should be constrained in systematic ways that yield testable empirical predictions.
Second, is a dual-systems framework appropriate?
When theorists invoke a dual-systems account of reasoning, one fundamental assumption is that the two systems compute the same function in two different ways: an initial, rapid system computes a heuristic response based on one or more cues, and a slower, deliberative system processes the same information in a more elaborate manner (Stanovich & West Reference Stanovich and West2000). The two systems rely on different algorithms to carry out the same cognitive task. But when H&M distinguish updating from reasoning, the goals of the two systems they posit differ: People update their perceptual models to maintain an accurate simulation of reality. In contrast, a person may engage in temporal reasoning to achieve many different goals, for example, planning for the future, reinterpreting the past, comprehending discourse, and understanding the sequence of a film. Because temporal updating and temporal reasoning are used for different purposes, invoking the dual-systems framework may be inappropriate.
Indeed, it is not clear to us why updating and reasoning are the only processes relevant to temporal cognition. Some tasks that require the representation of time do not require reasoning at all. Consider the task of event recall (Wang & Gennari Reference Wang and Gennari2019). The task requires an individual to recall events that comprise some temporal interval. For example, you might summarize your previous day as follows: “I had breakfast, worked on a project, taught a class, had a meeting, then had lunch with a friend….” The task requires individuals to remember and then to represent multiple events along a mental timeline. It does not concern temporal updating and it does not require reasoning, either, because responders need not infer any novel temporal relations while recalling events in memory. The act of remembering a temporal sequence seems fundamental to temporal thinking, but the dual-systems framework that H&M espouse has no place for it.
Hence, H&M must explain whether their account allows for cognitive processes that result in mental representations of temporally ordered events, even those that do not demand explicit temporal reasoning. The “intermediate developmental stage” (sect. 3, para. 1) to which they refer presents a broad challenge to the dual-systems framework. Children may struggle to retrieve temporal sequences, not because they revert to updating, but rather because of episodic memory retrieval failures (Prabhakar & Ghetti Reference Prabhakar and Ghetti2019). H&M should enumerate the specific pattern of errors predicted by reverting to the updating system. Perhaps a more accommodative framework, one that retains the division outlined by H&M, should specify the different processes relevant to temporal cognition (e.g., updating, recall, reasoning) as well as the various representational and computational constraints of each process (cf. Khemlani et al. Reference Khemlani, Harrison and Trafton2015).
In sum, H&M's distinction between temporal updating and reasoning is useful, so much so that it is worth refining, clarifying, and extending to address the two issues highlighted above.