Today is tomorrow’s yesterday: Children’s acquisition of deictic time words
Introduction
To learn their first words, children must rely primarily on the extra-linguistic context in which those words are used, since they are not yet able to understand the sentences in which the words are embedded. Consequently, children’s first words often label concrete referents that can be ostensively identified, like “mama,” “doggie,” and “cup.” Other words, however, are more difficult to learn through observation of the world and may require children to recruit their knowledge of the linguistic context in which those words are embedded (e.g., Gillette et al., 1999, Gleitman, 1990, Gleitman et al., 2005, Snedeker and Gleitman, 2004). For example, the meanings of deictic time words, such as “yesterday” and “tomorrow,” cannot be gleaned solely from extra-linguistic situations. These words are abstract and describe periods in time, which are difficult to reference ostensively. Further, due to their deictic functions, these words do not have fixed denotations and cannot be understood without information about the time at which they are uttered (Fillmore, 1997/1977): Tuesday’s “tomorrow” is different from Wednesday’s “tomorrow.” Acquiring words like these is one of the greatest challenges that English-learning children face, as evidenced by the massive gap between their first use of deictic time words around age 3 and their eventual mastery of adult-like meanings in elementary school (Ames, 1946, Busby Grant and Suddendorf, 2011, Harner, 1975, Harner, 1981). However, the process through which these words are ultimately acquired—and thus the roles of linguistic and referential context—remains mysterious. Here, as a case study of abstract word learning, we explore children’s gradual construction of deictic time word meanings between ages 3 and 8.
While many children produce words like “yesterday” and “tomorrow” as early as age 2 or 3 (Ames, 1946, Busby Grant and Suddendorf, 2011, Dale and Fenson, 1996), they do not use them as adults do for several subsequent years (Ames, 1946, Busby Grant and Suddendorf, 2011, Harner, 1975, Harner, 1981, Nelson, 1998, Weist, 1989, Weist et al., 1991). According to parental report, two thirds of 3-year-olds produce the word “yesterday,” but fewer than 20% use the word in adult-like ways; by age 5, more than 80% of children produce “yesterday,” but, still, fewer than 60% use it like adults (Busby Grant & Suddendorf, 2011). Children struggle not just with production but also with comprehension: When asked to name an event that occurred “yesterday” or one that will occur “tomorrow,” only about a quarter of 3-year-olds can provide reasonable answers (Busby and Suddendorf, 2005, Suddendorf, 2010). These difficulties persist even later in acquisition, as 5-year-olds can correctly generate an event from “yesterday” only 66% of the time, and an event that will occur “tomorrow” only 63% of the time (Busby & Suddendorf, 2005).
Although children differ from adults in how they use time words, it remains possible that they nevertheless use them systematically, and that they construct their meanings gradually and in stages over the first 6 or 7 years of life. Consider an anecdote: When 21-month-old Franny tried to remove dirty dishes from the dishwasher, her mother stopped her and said, “We can empty it tomorrow.” Upon hearing this, Franny ran to her bedroom, climbed under her blanket, closed her eyes, and after a brief delay returned to the kitchen to begin the chore. For Franny, “tomorrow” seemed to mean something like “after waking up.” Just a few months later, Franny began producing the word “yesterday,” but used it to refer not only to events that happened the previous day, but also to events that happened two days ago, five minutes ago, and even several months earlier. Productions like Franny’s are thought to be quite common (Ames, 1946, Friedman, 1990, Harner, 1981, Harner, 1982, Nelson, 1998, Weist, 1989) and suggest that although young children do not use deictic time words in adult-like ways, they may have partial knowledge of their meanings.
Critically, knowledge of partial word meanings may not have been detectable by the comprehension measures used in previous studies (i.e., parental report and event naming). For example, although adult English speakers may judge that Franny fails to use “yesterday” correctly, Franny may nonetheless know that “yesterday” refers to a period of time, and that it refers to periods prior to the time at which it is uttered. Further, even if Franny were to develop an adult-like meaning of “yesterday” and understand that it refers to a specific period exactly one day ago, she might not be able to name an event that occurred “yesterday.” The ability to associate time words with life events depends not only on knowledge of these words’ meanings, but also on the ability to recall, order, and anticipate events (e.g., a capacity for “mental time travel”; Suddendorf & Corballis, 2007). These abilities develop slowly (Busby and Suddendorf, 2005, Suddendorf et al., 2011, Schacter et al., 2007). These considerations suggest that other methods are required to probe children’s knowledge—or partial knowledge—of deictic time word meanings.
Understanding the nature of children’s early uses of deictic time words—and the partial word meanings they may implicate—could provide critical insight into the inductive hypotheses children make about these words’ meanings. While there has not been systematic study of children’s partial knowledge of deictic time words during the long delay between initial production and adult-like usage, there are hints that children may acquire information about different facets of their meanings independently, with some acquired before others. These facets include a word’s deictic status (e.g., “yesterday” is in the past; “tomorrow” is in the future), its sequential order relative to other time words (e.g., “next week” is a time after “tomorrow”), and its remoteness from the present (e.g., “yesterday” is exactly one day from today). For instance, 3-year-olds appear to understand that “yesterday” and “tomorrow” refer to a non-present time, without knowing that they refer specifically to the past and future, respectively (Harner, 1975). Further, children struggle to grasp the differing causal implications of events from “yesterday” vs. “tomorrow” on the present until at least age 5, also suggesting that their understanding of the distinction between past and future is incomplete (Busby & Suddendorf, 2010). Together, these results suggest that children may first learn that deictic time words label periods in time, without understanding much about their deictic past/future status, order, or remoteness. Furthermore, children’s over-extension errors within the past or future, like Franny’s use of “yesterday,” suggest that at some stage, children may understand a word’s deictic status without understanding its remoteness (e.g., Harner, 1981, Nelson, 1998).
One reason to think that children may acquire information about a word’s deictic status, order, and remoteness separately is that there is substantial variation in how these facets of time are expressed across languages. In English, for instance, all time words refer to either the past, present, or future. By contrast, in Urdu, “kal” refers to a period exactly one day from the present—whether in the future or the past—and thus does not encode deictic status but does encode temporal remoteness. Other languages include terms that encode degrees of temporal remoteness that are not lexicalized in English. For example, German’s “übermorgen” and Georgian’s “zeg” label a period that is in the future, much like English “tomorrow,” except that they refer to the period exactly two days away. Meanwhile, German’s “vorgestern” and Japanese’s “ototoi” pick out a time in the past, like English “yesterday,” except that it refers to the period exactly two days ago. The fact that deictic time words vary across languages according to factors like deictic status and remoteness is consistent with the idea that these facets of meaning are dissociable and may be learned independently by children.
The goal of the present study was to explore whether English-learning children have systematic but partial meanings of deictic time words during the long delay between their initial production of these words and eventual adult-like usage. Critically, the nature of children’s partial meanings—i.e., the developmental sequence in which information about deictic status, order, and remoteness are acquired—could constrain theories of the process through which these words are learned and the informational sources that children might draw upon. Broadly, there are two sources of information a child might use to learn the meanings of these terms: The events that time words refer to (e.g., a birthday party), and the linguistic context in which these words appear. As we describe below, these sources of information are differentially suited to supporting children’s inferences about deictic status, order, and remoteness.
Mappings between deictic time words and the events they are used to describe could plausibly help children learn the deictic status and remoteness of these words. For example, by noting whether deictic time words are used to describe events that are anticipated (e.g., a birthday party tomorrow/next week/etc.) or in the past (e.g., a birthday party yesterday/last week/etc.; Johnson, Foley, Suengas, & Raye, 1988) children could learn the deictic status of these words. Children could also generate hypotheses about the approximate temporal remoteness indicated by these words, e.g., by using the strength of a memory trace to estimate the remoteness of a party that occurred “last week” (Friedman, 2003, Hinrichs, 1970). From an understanding of deictic status and remoteness, children could then make inferences about relative order: e.g., last year is before yesterday because it refers to a more remote time in the past. Thus, if children were to rely primarily on event mappings, then they might acquire knowledge of deictic status and remoteness in tandem, and later infer information about order.
If children leverage the broader linguistic context to learn deictic time word meanings, they might exhibit a different developmental trajectory, in which knowledge of deictic status and/or order is constructed prior to knowledge of remoteness. To begin, even before children have learned anything about the meanings of deictic time words, they could use the linguistic context to infer that deictic time words belong to a common lexical class. For example, children might observe that deictic time words often appear in similar sentence frames (e.g., “The party [happened/will happen] [yesterday/tomorrow/last week/next year].”), or that they are often used in response to “When” questions, and from this infer that these words have similar kinds of meanings (Backscheider and Shatz, 1993, Shatz, 1993, Tare et al., 2008).
After grouping deictic time words into a common class, children could use other linguistic cues to make inferences about the specific semantic content of these words. For example, children could use their early knowledge of English tense markings (Brown, 1973, De Villiers and De Villiers, 1978, Harner, 1976, Weist et al., 1991) to infer whether deictic expressions refer to events in the past (e.g., “He danced last year”) or in the future (e.g., “He will dance tomorrow”). This process, in which grammatical cues are used to restrict hypotheses about meaning, is known as syntactic bootstrapping (Brown, 1957, Gleitman, 1990, Gleitman et al., 2005, Landau and Gleitman, 1985, Naigles, 1996). Further, children could also leverage cues from discourse structure to infer the relations among deictic time words. For example, contrastive uses of these words such as “The package isn’t coming tomorrow, it’s coming later, next week” could provide information about relative order. Moreover, since order-of-mention in discourse typically reflects temporal order (Jakobson, 1966), children could use input such as, “Bobby danced at his birthday party last year, but probably won’t dance at his friend’s party tomorrow,” to infer the sequential order of individual lexical items like “last year” and “tomorrow.” In sum, if children rely primarily on the linguistic context to learn deictic time word meanings, then they might acquire knowledge of deictic status (supported by syntactic bootstrapping) and order (supported by discourse structure) prior to knowledge of remoteness.
Here, to explore whether children might have systematic but partial meanings for deictic time words during the gap between initial production and later adult-like usage, we asked 3- to 8-year-old children to place these words onto a spatial timeline. As described below, this method allowed us to separately assess knowledge of these words’ deictic status (i.e., past vs. future), sequential order, and remoteness from the present.
Spatial scales have been used extensively to study children’s mental representation of number (e.g., Barth et al., 2009, Booth and Siegler, 2006, Ebersbach et al., 2008, Kolkman et al., 2013, Laski and Siegler, 2007, Moeller et al., 2009, Siegler and Booth, 2004, Siegler and Opfer, 2003, Slusser et al., 2013, Sullivan and Barner, 2014), and a smaller number of studies have used spatial scales to assess children’s understanding of temporal sequence (Busby Grant and Suddendorf, 2009, Friedman, 2000, Friedman, 2002, Friedman and Kemp, 1998, Hudson and Mayhew, 2011). However, most timeline studies involving preschoolers have explicitly avoided deictic time words, and, furthermore, have been limited in their ability to tease apart children’s understanding of the different semantic facets of these terms. For instance, all timeline studies of preschoolers have used scales depicting either the past or the future, but not both simultaneously (Busby Grant and Suddendorf, 2009, Friedman, 2000, Friedman, 2002, Friedman and Kemp, 1998). Thus these studies could not gauge children’s knowledge of deictic status, because children never had to decide whether events came from the past or the future. Furthermore, most timelines used to test preschoolers were divided categorically into distinct regions that represented broad periods of time, such as “a short time ago” or “a long time ago” (e.g., Busby Grant and Suddendorf, 2009, Friedman, 2002, Friedman and Kemp, 1998). Since terms placed inside the same region are not distinguished from each other, it is difficult and sometimes impossible to use categorical timelines to probe children’s knowledge of sequential order or remoteness.
We thus developed a new timeline task that allowed us to independently assess children’s knowledge of deictic status, sequential order, and remoteness. In our task, 3- to 8-year-old children and adult controls used colored pencils to mark where deictic time words (e.g., “yesterday”) and events (e.g., the participant’s last birthday) should go on horizontal timelines that extended, left-to-right, from the past (“when you were a baby”) to the future (“when you’ll be a grown-up”), with the present moment (“right now”) indicated by a dividing line between past and future. Knowledge of the deictic status of a word was assessed by its placement to the left or right of the midpoint, regardless of its placement relative to other words. Knowledge of sequential order was assessed by the ordering of words along the line (e.g., whether “last week” was placed before “yesterday”), ignoring their relation to the present and the distances between them. And knowledge of remoteness from the present was assessed by the spacing of terms along the line (e.g., the distance between “last year” and “now,” compared to the distance between “yesterday” and “now”). Finally, to confirm that timeline performance was a valid measure of children’s developing semantic knowledge, we investigated whether it correlated with children’s ability to answer verbal questions about the meanings of deictic time words (e.g., “Which will happen first: Tomorrow or next week?”).
Section snippets
Participants
Children from the greater San Diego, CA, (n = 93) and Berkeley, CA, (n = 25) areas participated in this experiment, along with 38 young adult controls from the UC San Diego Psychology Department subject pool. Data collection continued until we reached our target of sixteen children in each age category. Since recruited participants also participated in a related study on temporal gesture, this target sample size was based on past studies on children’s gesture (e.g., Sauter, Uttal, Alman,
Results
We report four analyses of the timeline data: First, we assessed comprehension of the deictic status, sequential order, and temporal remoteness of the deictic time words. Second, we determined the typical ages of acquisition of these facets of meaning, pinpointing the age at which the majority of children displayed adult-like comprehension of each facet. Third, we calculated the contingencies between adult-like knowledge of these three facets of meaning: i.e., the degree to which adult-like
Discussion
Our study investigated what children know about deictic time words during the long delay between when they begin producing these words, around age 3, and eventual adult-like usage in elementary school. Specifically, we asked whether children assign systematic, preliminary meanings to these words during this period, by using a timeline measure to independently assess children’s knowledge of three facets of their meanings: deictic status, order, and remoteness. Consistent with previous studies
Conclusion
Acquiring mature meanings for deictic time words can take children four years or more. Our findings suggest that, during the long delay between children’s initial production of these words and eventual adult-like usage, children construct systematic, partial meanings, including information about deictic status and order. One intriguing possibility is that these partial meanings are built through a gradual inductive process in which children construct an ordered, semantic domain for these words
Acknowledgments
We thank Sara Al Mughairy, Katherine Kimura, Miriam Rubenson, Karen Bejar, Sean Burdeaux, Kaitlyn Peck, Nestor Tulagan, Tilly Sirota, and Eren Fukuda for help with subject recruitment, data collection, and data entry. We are most grateful to the daycares and museums where data were collected, including the Reuben H. Fleet Science Center in San Diego, and to all the children and families who participated in this research. Thanks to Laura Wagner and two anonymous reviewers for constructive
References (72)
- et al.
Children’s mappings of large number words to numerosities
Cognitive Development
(2009) - et al.
Production of temporal terms by 3-, 4-, and 5-year-old children
Early Childhood Research Quarterly
(2011) - et al.
Recalling yesterday and predicting tomorrow
Cognitive Development
(2005) - et al.
Does learning to count involve a semantic induction?
Cognition
(2012) - et al.
The relationship between the shape of the mental number line and familiarity with numbers in 5-to 9-year old children: Evidence for a segmented linear model
Journal of Experimental Child Psychology
(2008) - et al.
The effects of elapsed time and retrieval on young children's judgments of the temporal distances of past events
Cognitive Development
(1998) - et al.
Human simulations of vocabulary learning
Cognition
(1999) - et al.
Children’s temporal judgments for autobiographical past and future events
Cognitive Development
(2011) - et al.
Early numerical development and the role of non-symbolic and symbolic skills
Learning and Instruction
(2013) - et al.
One, two, three, four, nothing more: An investigation of the conceptual sources of the verbal counting principles
Cognition
(2007)
Children’s early mental number line: Logarithmic or decomposed linear?
Journal of Experimental Child Psychology
Levels of number knowledge in early childhood
Journal of Experimental Child Psychology
The long and the short of it: On the nature and origin of functional overlap between representations of space and time
Cognition
The development of structural analogy in number-line estimation
Journal of Experimental Child Psychology
Learning the language of time: Children’s acquisition of duration words
Cognitive Psychology
Cross-cultural and developmental trends in graphic productions
Cognitive Psychology
Slow mapping: Color word learning as a gradual inductive process
Cognition
Time concepts in language and thought: Filling the Piagetian void between two to five years
Of metaphors and magnitudes: Explaining cognitive interactions between space, time, and number
Cortex
Children’s understanding of counting
Cognition
Children’s acquisition of the number words and the counting system
Cognitive Psychology
The development of the sense of time in the young child
Journal of Genetic Psychology
Children’s acquisition of the lexical domain of color
Developmental and individual differences in pure numerical estimation
Developmental Psychology
Linguistic determinism and the part of speech
The Journal of Abnormal and Social Psychology
A first language: The early stages
Towards an agenda for studying learning and instruction focusing on time-related competences in children
Educational Studies
Preschoolers begin to differentiate the times of events from throughout the lifespan
European Journal of Developmental Psychology
Young children’s ability to distinguish past and future changes in physical and mental states
British Journal of Developmental Psychology
Temporal distance: Remoteness distinctions in tense-aspect= systems
Linguistics
Tense and aspect systems
Lexical development norms for young children
Behavior Research Methods, Instruments, & Computers
Cited by (47)
Episodic thought in development: On the relation between memory and future thinking
2023, Developmental ReviewThe effects of cueing episodic future thinking on delay discounting in children, adolescents, and adults
2022, CognitionCitation Excerpt :However, we believe that it is unlikely that the children in our experiment did not understand the relative durations of the conventional time periods referred to in the tasks. Existing studies show that, by 8–9 years (the youngest age group in our experiment) children typically understand that there are objective differences in the durations of these time periods as a result of explicit teaching in school (McCormack, 2015; Tillman & Barner, 2015; Tillman, Marghetis, Barner, & Srinivasan, 2017). Our future time perception task was an attempt to establish the subjective differences in how far away future time points feel to the participant, and even adults show time compression (discriminating less well between distant time points) in this type of task (Kim & Zauberman, 2019).
The future is in front, to the right, or below: Development of spatial representations of time in three dimensions
2021, CognitionCitation Excerpt :Given these well-documented effects of culture and individual differences on the orientation of the horizontal/vertical and sagittal mental timelines, a key open question concerns the development of these mappings. Relatively few studies have investigated the developmental origins of spatiotemporal mappings, and these past studies have focused primarily on the origins of the horizontal/vertical mental timeline (Busby Grant & Suddendorf, 2009; Tillman, Marghetis, Barner, & Srinivasan, 2017; Tillman, Tulagan, Fukuda, & Barner, 2018; Tversky et al., 1991; but see Marghetis, Tillman, Srinivasan, & Barner, 2014; Burns, McCormack, Jaroslawska, O'Connor, & Caruso, 2019 for investigations of children's temporal gestures). One open-ended method that has been used to assess the development of the horizontal/vertical mental timeline is to ask children to place stickers on a sheet of paper to represent the relative order of multiple events (Tillman et al., 2018; Tversky et al., 1991).
Children's epistemic forecasting: The case of knowledge loss
2020, Journal of Experimental Child PsychologyCitation Excerpt :The 4-year-olds’ projected epistemic trajectory thus may reflect the salient points of the narrative. In addition, as noted, 4-year-olds might not have a precise idea of the spacing between today, tomorrow, and adulthood (Busby Grant & Suddendorf, 2011; Tillman et al., 2017, 2018; Zhang & Hudson, 2018). To the extent that not just 4-year-olds’ but also 6-year-olds’ understanding is still developing (Tillman et al., 2017), performance may undergo further changes with age.