First words and first memories

Cognition 116 (2010) 23–32

Contents lists available at ScienceDirect

Cognition journal homepage: www.elsevier.com/locate/COGNIT

First words and first memories Catriona M. Morrison *, Martin A. Conway Institute of Psychological Sciences, The Leeds Memory Group, University of Leeds, Leeds, LS2 9JT, England, United Kingdom

a r t i c l e

i n f o

Article history: Received 29 April 2009 Revised 8 March 2010 Accepted 10 March 2010

Keywords: Age of acquisition Childhood memories Episodic Autobiographical Concepts Amnesia

a b s t r a c t In two experiments autobiographical memories from childhood were recalled to cue words naming common objects, locations, activities and emotions. Participants recalled their earliest specific memory associated with each word and dated their age at the time of the remembered event. A striking and specific finding emerged: age of earliest memory was systematically later, by several months, than the age of acquisition of the word to which it was associated. This was the case for earlier and later acquired words, for all word types, and for younger as well as older adults. It is suggested that this systematic lag reflects the formation of conceptual knowledge that is abstracted from details represented in early episodic memories. It is not until such knowledge is formed that a word cue and the conceptual knowledge in long-term to which it corresponds, can be used to access specific episodic memories. The implications of this for understanding childhood amnesia and for theories of the development of autobiographical memory are considered. Ó 2010 Elsevier B.V. All rights reserved.

1. Introduction It is well known that adults can remember relatively few events from below the age of about 7 years, even fewer from below the age of 5 years, and it is thought none from the pre-verbal period of about 24 months of age and younger (Hayne, 2004; Howe & Courage, 1997; Howe, Courage, & Rooksby, 2009; Pillemer & White, 1989; Waldfogel, 1948;Wang, 2001; Wetzler & Sweeney, 1986 – see Bauer (2007), for a recent review). There are several different explanations for this period of childhood amnesia and most postulate some form of deficit, e.g., that memory, language, concepts, executive function, working memory, even the brain itself, have to attain some level of development before memories can be encoded and/or encoded in a retrievable form e.g., Newcombe, Lloyd, & Ratliff, 2007; Nelson, 1992. Quite feasibly all these factors contribute in some way to impaired adult recall from the age of seven and below. However, what is now clear is that children below the age of seven have many memories of specific experiences * Corresponding author. Tel.: +44 113 3 43 57 46. E-mail addresses: [email protected] [email protected] (M.A. Conway).

(C.M.

Morrison),

0010-0277/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.cognition.2010.03.011

and this may be true even of the pre-verbal period, so an encoding deficit account is not a tenable explanation of childhood amnesia (Bauer, 2004; Hayne, 2004; Nelson & Fivush, 2004). Instead, the evidence shows that the problem lies not in forming memories but in representing them in long-term memory in a way that renders them retrievable; we term this the retrievability hypothesis. How this is achieved by the developing infant/child is not known, although here we focus on one of the important factors: the development of language and in particular the development of vocabulary. We postulate that it is not until the developing child can name aspects of experience that they can begin to encode their memories in a way that will make them retrievable later in life. If this retrievability hypothesis is correct there should be a systematic relationship between the age of acquisition (AoA) of words and the age of earliest memories, the age at encoding (AaE), associated with those words. These two measures, AoA of words and AaE of memories, form the main variables of the present studies. AoA is a concept that has been shown to offer explanatory power for many key effects in lexical processing, for example, early-acquired words are easier to recognise (e.g., Bonin, Chalard, Meot, & Fayol, 2001; Morrison & Ellis,

24

C.M. Morrison, M.A. Conway / Cognition 116 (2010) 23–32

2000), quicker to name (e.g., Morrison, Hirsh, Chappell, & Ellis, 2002; Morrison, Hirsh, & Duggan, 2003), and have briefer gaze durations in sentence processing (Juhasz & Rayner, 2003, 2006). The AoA measure is usually a subjective, rated measure of word learning age. Adult raters are asked to make a judgement about the age at which they learnt certain words, typically along the seven-point scale introduced by Gilhooly and Logie (1980). Despite the fact that subjective ratings are widely used in the lexical processing literature (familiarity, imageability and concreteness are just some of the many measures that rely on subjective estimates), the validity of AoA has raised particular concern. However, there are several indicators that subjective measures of AoA are valid. For example, two studies (Chalard, Bonin, Meot, Boyer, & Fayol, 2003; Morrison, Chappell, & Ellis, 1997) compared subjective, adult ratings with objective measures of AoA from the study of children’s vocabulary development. Both concluded that objective AoA was the best predictor of subjective AoA. Several other studies have also compared children’s vocabulary knowledge with AoA ratings and concluded that they provide a valid measure of word learning age (Carroll & White, 1973; De Moor, Ghyselinck, & Brysbaert, 2000; Gilhooly & Gilhooly, 1980; Jorm, 1991). These thorough efforts to prove the validity of AoA give us confidence in the measure. Age at encoding has been widely used in the study of autobiographical memory. It has proved particularly useful in the study of memories across the lifespan and in identifying the lifespan retrieval curve (Conway, 1990; Conway & Pleydell-Pearce, 2000; Rubin, Wetzler, & Nebes, 1986 – see Conway & Williams (2008), for a recent review). The lifespan retrieval curve, in people aged about 35 years and older, has three main components: a recency component, the reminiscence bump, and the period of childhood amnesia. The recency portion of the curve shows good memory for recent experiences gradually decreasing as the retention interval lengthens, until that is the curve enters the period of the reminiscence bump. The reminiscence bump is marked by an increase in recall of autobiographical memories and other knowledge too. In its broadest terms the period of the reminiscence bump encompasses the two decades between the ages of 10– 30 years. However, many studies find the bump to occur in the period of 15–25 years of age (Conway & Rubin, 1993; Rubin, Rahhal, & Poon, 1998). The bump, which is one of the most stable findings in the study of human memory (Conway & Rubin, 1993) is identified solely by the AaE of memories which are virtually always supplied by the participant. The period of childhood amnesia is marked by a fall in the accessibility of memories and below the age of about 5 years by an exponential decline in memory retrievability. This too is one of the most stable findings in the study of autobiographical memory and is identified by AaE of memories and the number of memories that can be recalled (see Hayne (2004), for a review). Age at encoding is then one of the important variables in the study of autobiographical memory; however, although it produces reliable and stable data, AaE judgements are estimates rather than exact dates. Calendar time is rarely part of autobiographical memories which typically

are dated by inference and with reference to other memories or other known dates with which they are associated (Thompson, Skowronski, Larsen, & Betz, 1996). This is especially true of childhood memories because at the time of encoding few children would have knowledge of calendar dates and even fewer would be able to incorporate these into memories. Childhood memories are then dated with reference to other known dates such as a birthday, holiday, the birth of a sibling, or some other important family event (see Bauer (2007) for a review). Because of this estimates of AaE of childhood events are approximate and only rarely exact. Nonetheless, reliable patterns are found in AaE judgements, as outlined above, and accuracy is possible in terms of months and years rather than weeks, days, and smaller units of time (Bauer, 2007; Thompson et al., 1996). In the present research a high degree of accuracy of dating is not essential and that is because we are concerned with the relation between AaE of memories and AoA of words used to cue memory retrieval. Our expectation is that if there is a relation between language development, vocabulary acquisition, and memory then there should be a systematic relation between AaE of earliest memories and AoA of word cues. According to our retrievability hypothesis there should then be a reliable relation between AoA of words and AaE of earliest memories that can be elicited to those words. Consider, for example, an early-acquired word such as ball. Retrievability posits than many memories that feature the item ball can and most likely will be formed during the period of childhood amnesia. However, memories that can be accessed by the word ball must contain some representation of that word and/or the concept that the word activates. Memories that contain this knowledge will be accessible by ball in adulthood. Thus, the earliest memory retrievable by an adult to a word such as ball should in terms of AaE be close in time to the AoA of that word. It follows that in general the AoA of a word should be close to the AaE of the earliest memory retrievable to that word. It is this aspect of the retrievability hypothesis that we test here. 2. Experiment 1 In the experiments reported here we use the cue word method to investigate the association between AoA of words and the earliest memories (AaE) that can be retrieved to the words. Note that the word cues are taken from AoA norms and therefore their AoA is known. Participants are presented with words individually and required to recall the earliest memory they can that is associated with what is named by the word. They are instructed that the memory must be of a specific event that lasted minutes or hours and cannot be of more general knowledge of one’s life (see Conway & Pleydell-Pearce, 2000). 2.1. Method 2.1.1. Participants The participants were fifty undergraduate students at the University of Leeds, who took part in partial fulfilment

25

C.M. Morrison, M.A. Conway / Cognition 116 (2010) 23–32

of course requirements. There were 44 women and 6 men with a mean age of 19.96 years.

instructions went onto explain the 7-point rating scale: the first point on the scale represented a memory occurring at 0–2 years, the second point 3–4 years, the third point 5–6 years, the fourth point 7–8 years, the fifth point 9–10 years, the sixth point 11–12 years, and the seventh point 13+ years. This is the most commonly used scale to rate AoA (e.g., Morrison et al., 2003). There was a blank box beneath each word in which participants were required to provide a description of the memory; they were advised that this section was compulsory, and that, however brief the description, this was a key component of the study. Participants were also required to give a confidence rating for the memory and their age at the time by placing a mark on a 10 cm line which represented a continuum from ‘‘not at all confident” to ‘‘extremely confident”. Standard instructions concerning the specificity of the memory were also provided (see Conway & Bekerian, 1987).

2.1.2. Stimuli Fifty noun words from Morrison et al. (1997) were selected on the basis that they named items to which memories could be easily retrieved and were normally distributed on a core set of lexical measures: AoA, frequency, imageability and familiarity (see Appendix A). We selected words that in our judgment named distinct and easily identifiable concepts that most participants could relate to, ruling out items that participants might have difficulty with either because of their obscurity, e.g., armadillo, or their universality, e.g., chair. The words we selected are typical of the range of words used in lexical tasks (for a comparison, see stimuli reported in a recent paper by Catling & Johnston, 2009). In view of the fact that many common, high-frequency nouns are learnt in the preschool years, particularly imageable object names of the sort in the Morrison et al. database, this means necessarily that most of the words in our stimulus set will have been learnt in this narrow window of time. What is striking is that, despite there being differences of a matter of months between the acquisition of respective words, there is a marked AoA effect across many language tasks, as reviewed earlier.

3. Results and discussion Age at encoding (AaE) scores were derived by averaging across all participants’ ratings for each item. Confidence ratings were scored on a 7-point scale by measuring the point on the line at which the participant had marked their confidence level and assigning a value to two decimal places. For example, a mark placed precisely midway on the line was given a score of 3.50. AaE scores were transformed into a score in months for the purpose of direct comparison with AoA norms (Morrison et al., 1997). Table 1 shows the descriptive statistics of these ratings and the standard lexical variables. An important feature of the data is that AoA scores were on average 12 months lower than AaE scores. This difference was found to be highly reliable, t(49) = 9.81, p < .0001, and suggests that the age at which words are acquired is consistently earlier than the age of the earliest memories retrievable to those words. Table 2 shows the correlations between the variables. Note that frequency values were taken from the Celex Lexical Database (Baayen, Piepenbrock, & Gulikers, 1995) and the objective AoA measure was from children’s data

2.1.3. Procedure Participants were given a questionnaire to complete in their own time and return to the experimenter within seven days. The cover sheet explained the task and the rating scale. The instructions began: ‘‘We would like you to think about your earliest memories for a set of fairly common, everyday objects. We would like you to recall the first memory you have for each of a list of named objects, so we need you to dig deep into your memory and think about the first memory you have that is associated with each of the listed objects. We would like you to estimate how old you were at the time of each event in memory, and rate how confident you are that the memory and your estimate of your age at the time are accurate.” The

Table 1 Descriptive statistics for stimuli used in Experiment 1. AaE = age at encoding; AoA = age of acquisition.

Mean SD Range

Rated AaE (months)

Rated AoA (months)

Confidence (scales 1–7)

Imageability (scales 1–7)

Familiarity (scales 1–5)

Log frequency

64.81 13.85 39.7–95

52.69 13.71 23.4–84

4.57 .41 3.71–5.40

6.32 0.19 6–6.7

2.66 0.68 1.64–4.36

0.91 0.56 0–2.20

Table 2 Correlations between variables in Experiment 1.

AaE Rated AoA Objective AoA Confidence Imageability Familiarity * **

p < .05. p < .01.

AaE

Rated AoA

Objective AoA

1

.801** 1

.704** .791** 1

Confidence .680** .521** .388** 1

Imageability

Familiarity

.201 .247 .485** .205 1

.436** .312* .228 .275 .053 1

Frequency .310* .389** .302* .295* .157 .420**

26

C.M. Morrison, M.A. Conway / Cognition 116 (2010) 23–32

(Morrison et al., 1997). AaE correlated very highly with both measures of AoA, and also with confidence, familiarity, and to a lesser extent frequency. In order to examine relations between these variables two multiple regressions were conducted with AaE as the dependent variable. The first regression used the objective measure of AoA from children’s data (Morrison et al., 1997), and in the second we used a rated measure of AoA (Morrison et al., 1997; Analysis 2). Both measures of AoA were found to be reliable predictors of AaE and R2 = .748, F(5, 49) = 26.13, p < .0001, R2 = .768, F(5, 49) = 29.19, p < .0001, respectively. Table 3 shows AoA was the best predictor of AaE and concepts whose names are learnt early in life had earlier memories associated with them than concepts whose names are learnt later in life. Confidence was also a reliable predictor and higher confidence ratings were given to the later memories Familiarity was also significant and earlier memories were recalled to more familiar concepts. Imageability and frequency were not reliably related to AaE. These findings show a close relationship between the age at which a word is acquired, AoA, and the age of occurrence of the earliest memory, AaE, that can be recalled to a word. While the correlation between the two measures is strikingly high, they are clearly different in that the AaE is on average 12 months older than AoA. Hence a word is in a person’s vocabulary for approximately 1 year preceding the earliest memories that can be recalled by an adult. A possible concern about this conclusion is that our stimuli were, necessarily, skewed to the lower end of the AoA spectrum. This raises the question of whether the systematic lag maintains across the full AoA range. In order to address this we rank ordered the stimuli from earliest to latest. We took the 12 earliest and 12 latest words and looked at the difference between AoA and AaE. For the earliest words, the difference was 12.1 months; for the latest words the difference was 8.5 months. While this difference is reduced for the later words there is obviously a marked lag remaining for the latest-acquired words. Hence while the lag is larger for the later words, this analysis does not indicate that the systematic lag we found in these data is carried by the earliest-acquired words. This systematic relationship and reliable age difference between AoA and AaE suggests an important role for language in the representation of autobiographical memories and we consider this further below.

4. Experiment 2 This experiment used the same cue word procedure but with fewer cue words in order to collect more data on each

memory. Moreover the cue words were grouped into categories as we were interested in any potential differences between cue types. Such differences in cue types have often been observed in studies of autobiographical memory where retrieval times to more abstract cues, e.g., words naming emotions, are slower than retrieval times to words naming common objects and locations (see Conway (1990), for a review). Finally, in order to detect any individual differences we used two groups of participants, older and younger adults. 5. Method 5.1. Participants The young participants were 34 undergraduate students at the University of Durham, who took part in partial fulfilment of course requirements. There were 22 women and 12 men with a mean age of 20.3 years. The older adults were recruited through senior citizen groups and retirement homes and all 37 were unpaid volunteers. There were 26 women and 11 men with a mean age of 71.6 years. 5.2. Stimuli We used 16 concept names drawn from four categories which we have previously and successfully used to elicit autobiographical memories in our laboratory. The categories and their exemplars were: events (Christmas, birthday, dream, holiday), places (garden, kitchen, park, shop), negative emotions (embarrassed, frightened, angry, guilty) and positive emotions (successful, happy, in awe, excited). 5.3. Procedure Participants were given a questionnaire that explained the aim of the study was to improve understanding of our earliest memories: ‘‘You are asked to bring to mind your earliest memory of an experience that lasted for a period of no longer than minutes or hours. In other words, simply recalling ‘my first year at school’ or ‘when we went to X on holiday’ is not sufficient. Instead you must bring to mind specific images and details.” Beneath a description of the memory participants were asked to provide ratings (all on a scale of 1–5, where 1 = low and 5 = high) of: the emotional intensity of the remembered event, the vividness of the memory, the personal importance of the remembered event, and the extent to which they had rehearsed the memory. They were also asked to judge whether the memory was real: ‘‘the extent to which you actually remember

Table 3 Multiple regression analysis of data from Experiment 1, with AaE as the dependent variable, with objective AoA (Analysis 1) and rated AoA (Analysis 2). Analysis 1

Objective AoA Confidence Imageability Familiarity Frequency

Analysis 2 t

b .574 .450 .150 .196 .055

p 6.061 5.309 1.697 2.254 .630

.000 .000 .097 .029 .532

t

b Rated AoA Confidence Imageability Familiarity Frequency

.600 .342 .011 .201 .110

p 6.620 3.947 .138 2.421 1.302

.000 .000 .891 .020 .200

27

C.M. Morrison, M.A. Conway / Cognition 116 (2010) 23–32 Table 4 Descriptive statistics for stimuli used in Experiment 2. AaE = age of event; AoA = age of acquisition. Young

Rated AaE (months) Rated AoA (months) Vivdness (scales 1–5) Rehearsal (scales 1–5) Importance (scales 1–5) Emotion (scales 1–5) Realness

Old

Mean

SD

Range

Mean

SD

Range

68.1 60.3 3.24 2.51 2.87 3.07 3.97

9.29 .33 .21 .30 .25 .54 .27

56.3–89.7 28.8–83 2.85–3.64 2.2–3.2 2.48–3.41 2.26–4.12 3.47–4.45

77.4 63.7 3.23 2.53 2.79 2.94 3.95

15.3 4.8 .33 .26 .39 .48 .28

51.8–100.5 28.9–105.3 2.71–3.72 2.14–2.97 2.0–3.0 2.17–3.55 3.39–4.38

it compared to it being derived from family stories, etc.” They also indicated when the remembered event occurred by giving an AaE rated in months, that is, how many months old they were at the time of the remembered event. As AoA ratings were not available for all the cue words new AoA ratings were collected from independent groups.1 Frequency values were taken from the Celex Lexical Database (Baayen et al., 1995). 6. Results The data were analyzed by items and, separately, by subjects. 6.1. By-items analyses For the by-items correlational analyses means were calculated across items for all measures. Table 4 shows the means for both groups of participants on each of the rated dimensions. There were two main findings: the first is that, as in Experiment 1, rated AaE was significantly later than rated AoA, both in young (t(15) = 6.04, p<.0001) and older adults (t(15) = 3.25, p < .01). The fact that this effect is present in later adulthood demonstrates the stability of these ratings across the lifespan. Second, older adults tended to give later AaE ratings than young adults (t(15) = 4.07, p < .005), and later AoA ratings (t(15) = 4.53, p < .0001), although the correlations between the young and old ratings were very high (for AaE, r = .833, p < .0001; for AoA, r = .721, p < .005). However, the mean ratings by young and older adults for all other measures, vividness, rehearsal, importance and emotion, were almost identical and no reliable differences were observed. Table 5 shows the correlations for the young group. It can be seen from Table 5 that AaE correlated very strongly with AoA, such that earlier memories corresponded to early-learnt words and older memories to older acquired words. AaE also correlated highly with imageability (there are earlier memories for highly imageable concepts) and familiarity (earlier memories for highly familiar concepts). The correlation with frequency was weaker although still 1 The AoA ratings were collected by using the standard questionnaire (Morrison et al., 1997). This was given to Durham University students, carers and employees in the residential homes visited, and other older adults attending community centers. There were 94 respondents in all. Mean AoA ratings did not differ by age groups and the means over all respondents for each rated word were used in the analyses of this experiment.

reliable, with earlier memories for more frequent words; and there was also a significant correlation with realness, such that the earlier the memory the lower the realness rating. None of the other memory measures of vividness, rehearsal, importance, and emotion correlated reliably with AaE. Table 6 shows the correlations for the older participants’ data, and again the correlation between AaE and AoA was very strong; also again imageability, frequency and familiarity were significantly correlated with AaE. For this group two memory measures correlated with AaE: importance (later memories were more important) and emotion (later memories were more emotional). Multiple regression analyses were conducted with AaE as the dependent measure and using the variables that had significant simple correlations with AaE as predictors; hence for the young data there were five predictor variables, AoA, imageability, familiarity, frequency and realness. The overall equation was highly reliable, R2 = .94, F(5, 10) = 31.2, p < .0001, showing good predictive power in the combination of independent variables, see Table 7. The only variable that had a reliable independent effect in this equation was AoA; the other four independent variables failed to reach significance, although imageability was of borderline significance. In the multiple regression of the older participants’ data we used six predictor variables: AoA, imageability, familiarity, frequency, importance and emotion. Again, the overall equation was significant, R2 = .87, F(5, 9) = 10.34, p < .0001, and the results are shown in Table 8. The effect of AoA was of borderline significance; all other independent variables did not reach significance in this equation. 6.2. By-subjects analyses Subjects analysis allowed us to compare directly responses from the young and older adults. In fact, there were relatively few differences between young and old in their ratings to different word types. We averaged responses across each participant for each of the word types: event, location, negative emotion and positive emotion, and examined the ratings given to each measure in relation to age. When AaE and AoA ratings were compared, AaE ratings were consistently rated as later than AoA ratings (F(1, 34) = 165.7, p < .0001), and there was a difference in word-type (locations and events had earlier AaE and AoA than both negative emotions and positive emotions, F(3, 102) = 18.9, p < .0001). However, the between-subjects factor of group was not significant, F(1, 34) = 2.44, p > .1,

28

C.M. Morrison, M.A. Conway / Cognition 116 (2010) 23–32

Table 5 Correlations for young participants’ responses in Experiment 2. AoA AaE AoA Frequency Familiarity Imageability Vividness Rehearsal Importance Emotion Realness

Frequency

.854** 1

.602* .605** 1

Familiarity

Imageability

Vividness

Rehearsal

Importance

Emotion

Realness

.796** .607** .575*

.855** .703** .587* .729**

.008 .096 .094 .140 .244 1

.195 .249 .083 .206 .186 .540* 1

.221 .039 .088 .197 .484 .505* .726** 1

.401 .296 .328 .365 .703** .604* .625** .852** 1

.500* .474 .269 .513* .689** .616* .408 .693** .889**

1

1

1

Table 6 Correlations for older participants’ responses in Experiment 2. AoA AaE AoA Frequency Familiarity Imageability Vividness Rehearsal Importance Emotion Realness

Frequency

.838** 1

.675** .575* 1

Familiarity

Imageability

.601* .758** .575*

.862** .812** .587* .729**

1

1

.412 .384 .168 .394 .542* 1

Rehearsal

Importance

Emotion

Realness

.118 .073 .056 .018 .207 .773**

.567*

.516*

.483 .197 .394 .659** .853** .760**

.395 .198 .438 .630* .892** .692** .860**

.411 .396 .110 .465 .535* .894** .667** .793** .862**

1

1

1

1

Table 7 multiple regression analysis of young participants’ data from Experiment 2, with AaE as the dependent variable. b AOA Imageability Familiarity Frequency Realness

.732 .211 .166 .072 .058

t

Sig.

4.41 1.18 1.32 .688 .510

.001 .264 .215 .507 .621

Table 8 Multiple regression analysis of older participants’ data from Experiment 2, with AaE as the dependent variable. b AOA Imageability Frequency Familiarity Importance Emotion

Vividness

.532 .430 .284 .333 .021 .143

t

Sig.

2.192 1.564 1.758 1.617 .080 .554

.056 .152 .113 .140 .938 .593

and so the ratings overall did not differ significantly between young and old. Furthermore, there was no significant interaction between group and rating (AaE versus AoA) or between group and word-type, indicating consistent ratings between young and old for AaE and AoA and across word types. When young and old were compared, they differed significantly on only two measures: AaE ratings to events (young = 61.9 months, old = 75.21 months, t(68) = 2.48, p < .05); and AaE ratings to negative emotions (young = 72.1, old = 85.9, t(68) = 2.49, p < .05). So the age effect for AaE ratings mentioned above is carried by ratings for

events and negative emotions, not locations or positive emotions. These results are consistent with the by-items analyses that, while the AaE ratings differ for young and old, such that older adults give on average later ratings than young adults, the groups do not differ in vividness, rehearsal, importance and emotion ratings. 7. General discussion The two key findings were first, that AaE of earliest memories was systematically about 12 months later than the AoA of words cues, and second, that this was true of many of the memories and words in the two experiments in which the correlations between AaE and AoA were highly reliable, regardless of category of word and age of participant. Together we refer to this as the systematic lag between AoA of words and the AaE of the earliest memories retrievable to those words. This systematic lag may reflect the development of an interesting function of language, which is to isolate and label features/details of memories. Verbally labeling memories and memory content also has implications for accounts of childhood amnesia and we consider both verbal labeling and impaired memory for childhood next. 7.1. Words, concepts, and episodic memories According to one view the ability to form episodic memories may be present and functioning very early in development, even in the pre-verbal period (Conway, 2005, 2009; Singer & Conway, in preparation – see too 3Bauer (2007), for a recent review of childhood memory that reaches much the same conclusion). The content of these early episodic memories is sensory-perceptual-affective in nature and

C.M. Morrison, M.A. Conway / Cognition 116 (2010) 23–32

they may provide part of the basis for the emergence of early concepts, which can be abstracted from them. This essentially is a reformulation of a view originally proposed by Nelson (1974). Nelson argued that for the infant in the process of acquiring vocabulary, single words initially refer to events. For example, when the mother rolls the ball to the child and says ‘ball’ then the word ‘ball’ is taken by the infant to refer to the whole event and not just the object. In this way early episodic memories become associated with specific words. As memories accrue in which the same or similar words occur but the memories themselves differ, for example the ball rolled on the carpet versus the ball rolled on the grass, then the referent of the word narrows from whole episode to a detail of an episode. At some point the word itself must become attached to a conceptual, rather than episodic, representation in longterm memory. This shift from initial reliance on episodic memories to conceptual knowledge has been identified in adult learning too where it also appears to occur over a period measured in months, at least for complex sets of knowledge (Conway, Gardiner, Perfect, Anderson, & Cohen, 1997). It can occur more rapidly for simple lists of items (Dewhurst, Conway, & Brandt, 2009). The shift then from episodic memory to conceptual knowledge may be a general feature of the acquisition of new knowledge. It is of especial relevance that patients suffering from anterograde amnesia, a symptom of which is the inability to form new episodic memories, also typically have major impairments in acquiring new concepts (Conway & Fthenaki, 2000). The present findings suggest that the process of verbally labeling features of episodic memories may be a relatively lengthy one, at least during development. If a word is acquired on average 12 months before the earliest memory that can be retrieved to it, then it follows that the process of applying a word to a memory may take place over several months or longer. We suggest that this systematic lag occurs because the process of acquisition of new words (in infancy) also entails the formation of new concepts to which the word becomes associated in long-term memory. The lag between acquiring a word and later being able to use that word to access early memories may reflect a period during which the word itself is used to isolate memory details that then gradually come to form a conceptual representation of the named detail in memory. Thus, we suggest that in the present experiments it is not the words themselves that access the early memories but rather the conceptual representations that they activate (Anderson & Conway, 1993; Conway, 1990). It is these conceptual representations that provide the cues with which to access memories. In this sense it might be that conceptual knowledge comes to organize episodic memories – or at least comes to be the source of effective cues with which to access them, cf. Westmacott and Moscovitch (2003). We suggest that the systematic lag between AoA of words and AaE of earliest memories reflects a process of concept formation in which initially whole memories have single words attached to them, later detailed content across memories comes to be named by a word and, finally, that content is (somehow) abstracted, driven perhaps by socialization processes (Nelson & Fivush, 2004) and becomes the meaning of the word in long-term memory. At the same time

29

this new conceptual knowledge also becomes a means of accessing associated memories. The importance of conceptual knowledge in accessing early memories is further reflected in the finding that the AaE of earliest memories for named concepts corresponded much more closely with lexical features of the concept’s name (AoA, familiarity and imageability) than with mnemonic features such as the vividness of the memory or the importance of the remembered event to the individual. Interestingly, the findings of Experiment 2 also showed that words naming complex concepts, emotion labels (both positive and negative), had reliably later AoAs and AaEs than words naming simpler concepts (objects and events). This is important because it shows that the lag exists no matter what the AoA of the word used to cue the memory and, hence, the reliable correlations in both experiments. Thus, whether or not a word is acquired late or early in childhood there is a lag between AoA and the earliest memory that can be retrieved to it. This lends general support to our suggestion of a slow formation of conceptual knowledge that occurs during word acquisition and early word usage prior to the concept being available to provide effective cues to memory retrieval. 7.2. The conceptual basis of childhood amnesia Early episodic memories contain largely sensory-perceptual-affective knowledge and are accessible by cues that correspond to their content (Tulving & Thomson, 1973). By isolating and naming specific details in the content of memories a slow process of concept formation takes place2 and the emerging conceptual knowledge gradually comes to organize memories. One straightforward implication that follows from this is that the period of childhood amnesia, impaired recall of memories from below about 7 years of age (Waldfogel, 1948) and virtually no memories from the pre-verbal period (see Hayne (2004), for a recent review), reflects the gradual emergence of conceptual knowledge over this period of development. Without the conceptual knowledge early episodic memories can only be accessed with cues that correspond in some way with their content. Presumably as the retention interval increases the probability of generating effective cues to these episodic memories decreases, in line perhaps with other endogenous and environmental changes that make the generation of effective cues increasingly less likely. In the older child and in the adult such cues can be found in conceptual knowledge which links directly to episodic memories. Indeed, as Tulving (1983) originally pointed out (adult) autobiographical memories contain conceptual knowledge. Moreover, in our account of autobiographical memory (Conway, 2005; Conway & Pleydell-Pearce, 2000) autobiographical memories are defined as episodic memories plus conceptual organizing context (see Conway (2009), for review). Episodic memories from early childhood are not accessible by the older child and adult because there is no 2 We do not claim that this is the only process of concept formation. Quite clearly there are others and in the adult perhaps a range of such processes. Our point is that this is an important early way in which concepts can be formed.

30

C.M. Morrison, M.A. Conway / Cognition 116 (2010) 23–32

or little conceptual knowledge with which to access them. Representing experience in such a way as to create a cue that could activate the content of these early sensory-perceptual-affective episodic memories is probably not possible for adults. Nevertheless, the intriguing possibility remains that such memories may endure in long-term memory, available but inaccessible, and what influence if any they continue to have on adult cognition is not known (see Singer & Conway, in preparation, for further discussion). On the other hand they may, of course, be lost from the system entirely. 7.3. The development of autobiographical memory We suggest then that one of the causes of childhood amnesia may be lack of conceptual knowledge with which to organize and access sensory-perceptual episodic memories. However, a more common explanation of childhood amnesia posits that the ability to form episodic memories has a slow developmental course and the reason why adults cannot remember much below the age of 5 years is that no, or few, enduring memories were formed in this period. We refer to this as the deficit theory. One influential form of deficit theory was proposed by Nelson (1992) who argued that prior to developing the ability to form episodic memories the child first has to develop schematic knowledge. It is schema of everyday activities and routines that support the representation of specific experiences in long-term memory. A more recent version of deficit theory has been formulated by Newcombe et al. (2007). According to this cognitive neuroscience approach the brain areas, in the medial temporal lobe memory system, that subserve the formation of episodic memories reach functionality over the first 2–3 years of life and this is the reason why adults cannot remember this period – no memories are formed. According this view it is not until semantic knowledge is formed in long-term memory that ability to generate and retain episodic memories begins to emerge. Indeed, Newcombe et al., argue that the neonate and preverbal infant are similar in their memory abilities to anterograde amnesics who cannot form new memories. Whether the medial temporal lobe is in fact so undeveloped at birth as to be unable to support the formation of episodic memories is, however, debatable. Moreover there is a wide range of evidence that children below the age of 3 years have memories and even some evidence that very young infants, only a few months old, show some evidence of episodic-specific memory (Rovee-Collier, 1996). Very importantly infants and children, unlike anterograde amnesics, show powerful signs of learning and acquisition of knowledge. Indeed, as we argued above it may be the case that episodic memory is essential to knowledge acquisition and not the other way round. On the other hand once conceptual knowledge is formed it organizes episodic memories and most importantly of all brings about the creation of autobiographical memory. An important distinction rarely made in accounts of the development of human memory is that between episodic memory and autobiographical memory. By our view episodic memory is a species-general memory facility and constitutes a fundamental learning mechanism (Conway,

1996, 2005, 2009; Conway & Piolino, 2009; Conway & Pleydell-Pearce, 2000). However, with the formation of extensive and elaborate conceptual knowledge, especially about one’s own life and self, access to sensory-perceptual episodic memories is rapidly lost, over periods measured in hours and days and very few can be retained in an accessible form. Language and conceptual knowledge, two distinctly human abilities, are what allow us to maintain access to episodic memories that themselves become linked into the complex knowledge base that is human long-term memory. Arguably, one of if not the most important conceptual development is the emergence of a conceptual distinction between the ‘me’ and ‘I’ (James, 1896). Developmentally, this occurs at about the age of 2 years and constitutes what Howe and Courage, (1997) termed the emergence of the cognitive self. These authors view this as the starting point for autobiographical memory (see Howe et al., 2009, chap, for recent discussion). Autobiographical memory is then episodic memory embedded in and organized by conceptual knowledge of the self and the world. The neonate and young infant rather than being functionally similar to anterograde anmesics are, by this view, more like a being that lacks a cognitive self and other conceptual knowledge. Without these structures in long-term memory access to sensoryperceptual episodic memories cannot be maintained. Indeed in the adult episodic memories will be encoded and even contain conceptual knowledge and it is this that preserves access for most memories – this is the retrievability hypothesis. Developmentally then the lag between AoA of words and AaE of memories retrievable to those words reflects this slow emergence of the integration of episodic memories with conceptual knowledge (named by words) or, by our view which we share with Howe and colleagues, the emergence of autobiographical memory. 8. Conclusion We have found a striking and highly reliable systematic lag between the age of acquisition of words and the age of earliest memories retrievable to those words. Earliest memories are several months older than the age of acquisition of words used to cue them. It is suggested that this systematic lag reflects a process of concept formation in which knowledge named by a word is isolated in early episodic memories and comes to form in memory the conceptual representation of a word and, simultaneously, a means of accessing memories. Lack of conceptual knowledge may explain childhood amnesia for the preverbal years and the slow emergence, through socio-cultural processes (Nelson & Fivush, 2004), of more extensive and organized conceptual knowledge may explain the gradual increase in memory accessibility over the early years of childhood. Acknowledgements Martin A. Conway was supported by the award of a Professorial Fellowship from the Economic and Social

C.M. Morrison, M.A. Conway / Cognition 116 (2010) 23–32

Research Council (ESRC), RES-051-27-0127 of the United Kingdom and he thanks the ESRC for this support. Appendix A. used in Experiment 1, taken from Morrison et al. (1997)

Axe Bell Button Cactus Church Cigarette Clown Cowboy Crab Crown Dice Doctor Dragon Dress Flag Flute Goat Gorilla Guitar Gun Harp Heart Jelly Jigsaw Ladybird Lamp Lemon Lorry Mermaid Microphone Mushroom Nun Onion Parachute Peacock Penguin Pumpkin Queen Sandwich Screwdriver Snail Star Strawberry Swing Tie Tiger Trumpet Vase Whale Yo-yo

AoA (months)

Imageability (scales 1–7)

Familiarity (scales 1–5)

56.4 40.8 39.6 78 44.4 66 38.4 60 61.2 45.6 60 42 49.2 23.4 55.2 72 46.8 51.6 55.2 54 68.4 48 34.2 38.4 36 56.4 50.4 40.8 61.2 78 56.4 74.4 49.2 84 67.2 44.4 63.6 36 34.2 72 51.6 36 44.4 32.4 46.8 46.8 63.6 70.8 58.8 50.4

6.2 6.6 6.4 6.3 6.5 6.25 6.7 6.3 6.4 6.4 6.65 6.25 6.6 6.1 6.35 6.1 6.3 6.1 6.35 6.5 6 6.7 6 6.25 6.5 6 6.2 6.3 6.3 6.1 6.15 6.2 6.2 6.3 6.25 6.55 6.25 6.25 6.45 6 6.25 6.5 6.6 6.3 6.1 6.6 6.4 6.55 6.35 6.2

2.14 2.5 4.09 2.7 3.09 3.86 2.09 2.1 2.55 1.68 3 3.65 2.35 3.14 2.22 1.91 2 1.64 3 2 1.68 3.09 3 3 3 3.73 2.95 3.41 2.05 2.85 3.2 2.4 3.95 2.15 1.91 1.86 1.77 3.05 4.36 2.73 2.45 3.09 2.77 2.27 2.91 1.77 2.05 2.5 3.15 2.15

31

References Anderson, S. A., & Conway, M. A. (1993). Investigating the structure of autobiographical memories. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19, 1178–1196. Baayen, R. H., Piepenbrock, R., & Gulikers, L. (1995). The CELEX Lexical Database (Release 2). Philadelphia, PA: Linguistic Data Consortium, University of Pennsylvania [Distributor]. Bauer, P. J. (2004). New developments in the study of infant memory. In D. M. Teti (Ed.), Blackwell handbook of research methods in developmental science (pp. 467–488). Blackwell. Bauer, P. J. (2007). Remembering the times of our lives: Memory in infancy and beyond. Mahwah, NJ: Lawrence Erlbaum Associates, Publishers. Bonin, P., Chalard, M., Meot, A., & Fayol, M. (2001). Age-of-acquisition and word frequency in the lexical decision task: Further evidence from the French language. Cahiers de Psychologie Cognitive/Current Psychology of Cognition, 20, 401–443. Carroll, J. B., & White, M. N. (1973). Word frequency and age-ofacquisition as determiners of picture-naming latency. Quarterly Journal of Experimental Psychology, 25, 85–95. Catling, J. C., & Johnston, R. A. (2009). The varying effects of age of acquisition. Quarterly Journal of Experimental Psychology, 62, 50–62. Chalard, M., Bonin, P., Meot, A., Boyer, B., & Fayol, M. (2003). Objective age-of-acquisition (AoA) norms for a set of 230 object names in French: Relationships with psycholinguistic variables, the English data from Morrison et al. (1997), and naming latencies. European Journal of Cognitive Psychology, 15, 209–245. Conway, M. A. (1990). Autobiographical memory and conceptual representation. Journal of Experimental Psychology: Learning,Memory, and Cognition, 16(5), 799–812. Conway, M. A. (1996). What do memories correspond to? Behavioural Brain Research, 19(2), 195–196. Conway, M. A. (2005). Memory and the self. Journal of Memory and Language, 53(4), 594–628. Conway, M. A. (2009). Episodic memory and the brain. Neuropsychologia, 47, 2305–2313. Conway, M. A., & Bekerian, D. A. (1987). Organization in autobiographical memory. Memory & Cognition, 15(2), 119–132. Conway, M. A., & Fthenaki, A. (2000). Disruption and loss of autobiographical memory. In L. S. Cermak (Ed.), Handbook of neuropsychology, Memory and its disorders (2nd ed., pp. 281–312). Amsterdam: Elsevier. Conway, M. A., Gardiner, J. M., Perfect, T. J., Anderson, S. J., & Cohen, G. M. (1997). Changes in memory awareness during learning: The acquisition of knowledge by psychology undergraduates. Journal of Experimental Psychology: General, 126(4), 1–21. Conway, M. A., & Piolino, P. (2009). Tous les rouages de notre identite. La Recherche, 432. Conway, M. A., & Pleydell-Pearce, C. W. (2000). The construction of autobiographical memories in the self memory system. Psychological Review, 107, 261–288. Conway, M. A., & Rubin, D. C. (1993). The structure of autobiographical memory. In A. F. Collins, S. E. Gathercole, M. A. Conway, & P. E. Morris (Eds.), Theories of memory. Hillsdale: Erlbaum. Conway, M. A., & Williams, H. L. (2008). Autobiographical memory. In H. L. Roediger, III (Ed.), Cognitive Psychology of Memory: Vol 2. Learning and Memory: A Comprehensive Reference (pp. 893–909). Oxford: Elsevier. De Moor, W., Ghyselinck, M., & Brysbaert, M. (2000). A validation study of the age-of-acquisition norms collected by Ghyselinck, De Moor and Brysbaert. Psychologica Belgica, 40, 99–114. Dewhurst, S. A., Conway, M. A., & Brandt, K. R. (2009). Tracking the R-to-K shift: Changes in memory awareness across repeated tests. Applied Cognitive Psychology, 23, 848–859. Gilhooly, K. J., & Gilhooly, M. L. (1980). The validity of age-of-acquisition ratings. British Journal of Psychology, 71, 105–110. Gilhooly, K. J., & Logie, R. H. (1980). Age-of-acquisition, imagery, concreteness, familiarity, and ambiguity measures for 1,944 words. Behavior Research Methods & Instrumentation, 12, 395–427. Hayne, H. (2004). Infant memory development: Implications for childhood amnesia. Developmental Review, 24, 33–73. Howe, M. L., & Courage, M. L. (1997). The emergence and early development of autobiographical memory. Psychological Review, 104, 499–523. Howe, M. L., Courage, M. L., & Rooksby, M. (2009). The genesis and development of autobiographical memory. In M. L. Courage & N. Cowan (Eds.), The development of memory in infancy and childhood (pp. 177–196). Hove, UK: Psychology Press. James, W. (1896). Principles of psychology. New York: Holt.

32

C.M. Morrison, M.A. Conway / Cognition 116 (2010) 23–32

Jorm, A. F. (1991). The validity of word age-of-acquisition ratings: A longitudinal study of a child’s word knowledge. British Journal of Developmental Psychology, 9, 437–444. Juhasz, B. J., & Rayner, K. (2003). Investigating the effects of a set of intercorrelated variables on eye fixation durations in reading. Journal of Experimental Psychology: Learning, Memory, and Cognition, 29, 1312–1318. Juhasz, B. J., & Rayner, K. (2006). The role of age-of-acquisition and word frequency in reading: Evidence from eye fixation durations. Visual Cognition, 13, 846–863. Morrison, C. M., Chappell, T. D., & Ellis, A. W. (1997). Age of acquisition norms for a large set of object names and their relation to adult estimates and other variables. Quarterly Journal of Experimental Psychology, 50(A), 528–559. Morrison, C. M., & Ellis, A. W. (2000). Real age of acquisition effects in word naming and lexical decision. British Journal of Psychology, 91, 167–180. Morrison, C. M., Hirsh, K. W., Chappell, T., & Ellis, A. W. (2002). Age and age of acquisition: An evaluation of the cumulative frequency hypothesis. European Journal of Cognitive Psychology, 14, 435– 459. Morrison, C. M., Hirsh, K. W., & Duggan, G. B. (2003). Age of acquisition, ageing, and verb production: Normative and experimental data. Quarterly Journal of Experimental Psychology, 56(A), 705–730. Nelson, K. (1974). Concept, word, and sentence: Interrelations in acquisition and development. Psychological Review, 81(4), 267–285. Nelson, K. (1992). The psychological and social origins of autobiographical memory. Psychological Science, 4, 7–14. Nelson, K., & Fivush, R. (2004). The emergence of autobiographical memory: A social cultural developmental theory. Psychological Review, 111, 486–511. Newcombe, N. S., Lloyd, M. E., & Ratliff, K. R. (2007). Development of episodic and autobiographical memory: A cognitive neuroscience

perspective. In R. V. Kail (Ed.). Advances in child development and behavior (Vol. 35, pp. 37–85). San Diego, CA: Elsevier. Pillemer, D. B., & White, S. H. (1989). Childhood events recalled by children and adults. In H. W. Reese (Ed.). Advances in child development and behavior (Vol. 21, pp. 297–340). Orlando, FL: Academic Press. Rovee-Collier, C. (1996). Shifting the focus from what to why. Infant Behavior and Development, 19(4), 385–400. Rubin, D. C., Rahhal, T. A., & Poon, L. W. (1998). Things learned in early adulthood are remembered best. Memory & Cognition, 26, 3–19. Rubin, D. C., Wetzler, S. E., & Nebes, R. D. (1986). Autobiographical memory across the adult lifespan. In D. C. Rubin (Ed.), Autobiographical memory (pp. 202–221). Cambridge: Cambridge University Press. Singer, J.A., Conway, M.A. (in preparation). Memory’s Duality: Enactive Memory in the Self-Memory System, Loewald Revisited Thompson, C. P., Skowronski, J. J., Larsen, S. F., & Betz, A. L. (1996). Autobiographical memory: Remembering what and remembering when. Mahwah, NJ: Lawrence Erlbaum Associates Inc.. Tulving, E. (1983). Elements of episodic memory. New York: Oxford University Press. Tulving, E., & Thomson, D. M. (1973). Encoding specificity and retrieval processes in episodic memory. Psychological Review, 80, 352–373. Waldfogel, S. (1948). The frequency and affective character of childhood memories. Psychological Monographs, 62(4) [Whole No. 291]. Wang, Q. (2001). Culture effects on adults’ earliest childhood recollection and self-description: Implications for the relation between memory and the self. Journal of Personality and Social Psychology, 81(2), 220–233. Westmacott, R., & Moscovitch, M. (2003). The contribution of autobiographical significance to semantic memory. Memory & Cognition, 31(5), 761–774. Wetzler, S. E., & Sweeney, J. A. (1986). Childhood amnesia: An empirical demonstration. In D. C. Rubin (Ed.), Autobiographical memory (pp. 191–201). Cambridge: CUP.