Document Reading and Writing

Unlock access to all the studying documents.

View Full Document

Vol.:(0123456789)

Reading and Writing

https://doi.org/10.1007/s11145-020-10083-9

1 3

The origin ofcentrality deficit intext memory

andcomprehension bypoor comprehenders: athink‑aloud

study

MenahemYeari1 · ShirleyLantin1

Abstract

The present study employed a think-aloud method to explore the origin of a cen-

trality deficit (i.e., poor recall of central ideas) found in poor comprehenders (PC).

Moreover, utilizing the diverse think-aloud responses, we examined the overall qual-

ity of text processing employed by PC during reading, in order to shed more light on

the cognitive underpinnings underlying their poor comprehension and memory after

reading. To address these goals, adolescents with good and poor comprehension,

matched on reading (decoding) skills, were asked to state aloud whatever comes

to their mind during the reading of two expository texts. After reading, the partici-

pants freely recalled text ideas and answered multiple-choice questions on the texts.

Results indicated that PC exhibited lower performance than good comprehenders

(GC) on the recall and comprehension tasks. The think-aloud protocols indicated

that PC generated fewer responses than GC that reflect high-level, deep text process-

ing, and more responses that reflect low-level, surface text processing. Furthermore,

compared to GC, PC reinstated fewer prior text ideas, with this reduction being sig-

nificantly greater for central than for peripheral ideas. Finally, the proportions of

deep processing responses in general were positively associated with participants’

performance on recall and comprehension tasks. These findings suggest that PC

exhibit poor text comprehension and memory, particularly of central ideas, because

they construct a low-quality, poorly-connected text representation during reading,

and produce fewer, less-elaborated retrieval cues for subsequent text comprehension

and memory. This explanation is further illuminated in the context of previous find-

ings and theoretical accounts.

Keywords Centrality deficit· Main ideas· Poor comprehenders· Reading

comprehension· Text recall· Think-aloud

* Menahem Yeari

Menahem.Yeari@biu.ac.il

1 School ofEducation, Bar-Ilan University, Ramat-Gan52900, Israel

M.Yeari, S.Lantin

1 3

Introduction

Poor comprehenders (PC), also referred to as individuals with specific read–

ing comprehension deficit, are readers who experience significant difficulties in

reading comprehension despite adequate reading ability (i.e., word decoding)

and normal (non-verbal) intelligence (Cain & Oakhill, 2007; Hulme & Snowl–

ing, 2011; Snowling & Hulme, 2012). In efforts to explore the sources of this

comprehension disorder, various language and cognitive skills, such as vocabu–

lary knowledge, inference generation, monitoring, and working memory, have

been examined in PC (for a review, see Cain & Oakhill, 2007). Surprisingly, their

ability to identify, process, and remember central (main) ideas, which are impor–

tant for the overall comprehension of a text, have hardly been examined in PC. A

recent study by Yeari and Lev (2020), which examined several types of central–

ity skills in PC, demonstrated their difficulty in recalling central ideas that they

had successfully identified, attended, and recognized during and after reading.

The present study was designed to illuminate the causes underlying this specific

retrieval difficulty, using a think-aloud method. In addition, utilizing the diverse

think-aloud responses, we examined the nature and quality of online text process–

ing employed by PC during reading and its effect on oﬄine poor comprehension

and memory they exhibit after reading.

Centrality eect andfidecit

Identifying central ideas during text processing and allocating sufficient cogni–

tive resources to process and encode these ideas in long-term memory is a crucial

skill for successful comprehension and learning of texts (Gersten, Fuchs, Wil–

liams, & Baker, 2001; Jitendra, Kay, & Xin, 2000; Meyer, Brandt, & Bluth, 1980;

Stevens, Slavin, & Farnish, 1991). Consistent with this notion, numerous stud–

ies have found that skilled readers recall more central than peripheral ideas after

reading (Britton, Meyer, Hodge, & Glynn, 1980; Britton, etal., 1986; Brown &

Smiley, 1977; Cirilo & Foss, 1980; Hyönä & Niemi, 1990; Keenan & Brown,

1984; Kintsch, Kozminsky, Streby, McKoon, & Keenan, 1975; Meyer, 1975;

Thorndyke, 1977; Yeari & Lev, 2020; Yeari etal., 2015, 2017; Yeari, Vakil, Schi–

fer, & Schiff, 2019). This superior recall of central ideas was referred to as cen–

trality effect (Miller & Keenan, 2009).

Interestingly, a centrality effect has been observed in disabled readers as well,

albeit in a weaker manner than in skilled readers. Specifically, disabled readers

recalled significantly fewer central ideas than their skilled peers, whereas no or

a lesser reduction has been observed in recalling peripheral ideas. This so-called

centrality deficit (Miller & Keenan, 2009), i.e., inferior recall of central ideas, was

found with poor readers (i.e., decoders) (Eamon, 1978; Miller & Keenan, 2009;

Smiley, Oakley, Worthen, Campione, & Brown, 1977), individuals with attention

disorder (Miller etal., 2013; Yeari etal., 2019), and individuals with unspecified

learning disabilities (Carlisle, 1999; Curran, Kintsch, & Hedberg, 1996; Hansen,

1 3

The origin ofcentrality deficit intext memory andcomprehension…

1978; Worden & Nakamura, 1982). A centrality deficit was also found in readers

who exhibited low performance in reading comprehension tests, whose decoding

skills had not been assessed (Drum, 1985; Winograd, 1984).

Centrality skills offipoor comprehenders

To the best of our knowledge, only two studies have examined the presence of cen-

trality deficit in PC, who had adequate decoding skills and no other learning dis-

abilities or attention disorder. In one study, Yuill and Joscelyne (1988) examined

the centrality effect after reading in 7–8 year-old good and poor comprehenders,

who were matched on reading skills and vocabulary knowledge. In this study, which

was primarily designed to examine the contribution of organizational cues to story

comprehension, participants read eight short stories accompanied with integrative

or non-integrative title and pictures. Each story included three central ideas—set-

ting/problem, initiating event, and direct consequence—and seven peripheral ideas,

defined by the researchers’ analysis of story grammar (Stein & Glenn, 1979). Par–

ticipants were asked to recall freely as much as they could remember from the story,

and answer comprehension questions after each story. Yuill and Joscelyne did not

find any difference in overall recall and the size of centrality effect between the two

groups. Both good and poor comprehenders recalled, to the same extent, more cen-

tral ideas than peripheral ones. They concluded that poor comprehenders are able to

discriminate between central and peripheral ideas, and prefer the processing of the

former during reading, despite their poor understanding of the story (as indicated by

their low scores on the comprehension questions).

In a recent study, Yeari and Lev (2020) examined several centrality skills of

adults with poor reading comprehension, who have normal word reading skills and

non-verbal intelligence. Specifically, they compared the extent that poor and good

comprehenders identify, attend, store, and retrieve central and peripheral ideas dur–

ing and after reading expository texts. To do so, participants read three texts, whose

ideas’ centrality level pre-estimated by human raters, while their eye-movements

were monitored during reading. After reading, participants were asked to recall,

recognize, and rate the centrality level of text ideas. Yeari and Lev found that PC

recalled significantly fewer central ideas than good comprehenders (GC) (along

with comparable recall of peripheral ideas; i.e., a centrality deficit), whereas no dif-

ference was found between the groups in (re)reading times and frequency and cor–

rect recognition proportions of central and peripheral ideas. A milder difficulty was

found for PC in estimating the centrality level of text ideas, which was more difficult

to interpret. Yeari and Lev concluded that PC have a specific difficulty in retrieving

central ideas which they had successfully attended and stored in long-term memory.

They suggested that this retrieval difficulty may be an outcome of their poor integra-

tion and inferential skills (Bowyer-Crane & Snowling, 2005; Cain & Oakhill, 1999;

Cain, Oakhill, Barnes, & Bryant, 2001; Oakhill, 1982, 1984; Yeari, Elentok, etal.

2017), particularly their ability to establish intra-text connections during reading and

construct retrieval cues for text recall after reading.

M.Yeari, S.Lantin

1 3

Based on theories of text processing and memory (Kintsch, 1998; Myers &

O’Brien, 1998; Singer & Kintsch, 2001), Yeari and Lev (2020) explained that a

retrieval of a text idea is facilitated or “cued” by the retrieval of other text ideas

which are conceptually connected with it, through spreading activation process that

occurs in the mental (network) representation of the text. Central ideas, which are

theoretically and empirically connected with more text ideas than peripheral ones

(Trabasso & van den Broek, 1985; Trabasso, van den Broek, & Suh, 1989; van den

Broek & Trabasso, 1986), benefit from more retrieval cues and are thus retrieved

to a greater extent than peripheral ideas. PC, who have exhibited low performance

in answering comprehension questions that require the integration of multiple text

ideas (Cain & Oakhill, 1999; Oakhill, 1982, 1984), presumably form fewer inter–

connections between text ideas during reading, and consequently benefit from fewer

retrieval cues for text ideas, particularly central ones, during text recall after reading.

Therefore, they have more difficulties than their peers in retrieving central ideas that

are available in their long-term memory. The present study was designed to examine

this hypothesis more directly, by employing a think-aloud procedure.

Think‑aloud byfipoor comprehenders

Several studies have used a think-aloud method to explore the nature of online text

processing by PC (Denton etal., 2015; Janssen, Braaksma, & Rijlaarsdam, 2006;

Laing & Kamhi, 2002; Magliano & Millis, 2003; Seipel, Carlson, & Clinton, 2017).

In these studies, poor and good comprehenders were instructed to say aloud what-

ever comes to their mind following reading or listening to sequential text segments.

This procedure aimed to uncover immediate, spontaneous comprehension processes

that take place during text processing, when comprehenders attempt to comprehend

and remember texts. Comparing the quality of text processing by poor and good

comprehenders, these studies were generally consistent in demonstrating that GC

generate more responses than PC which reflect high-level, deep text processing,

such as connecting text ideas, inferring from prior knowledge, and monitoring com-

prehension (see Discussion for a finer comparison of the results obtained in these

studies). Note, however, that different assessments were used to diagnose PC in

these studies (e.g., teachers’ rating, standardized tests, national tests), and none of

them tested and/or compared the reading abilities of poor and good comprehend-

ers (Denton etal., 2015, used word reading scores as exclusion criterion; Laing &

Kamhi, 2002, examined PC who received lower scores than GC in a word-reading

test). Moreover, these studies did not examine performance in oﬄine tasks after

text processing (Denton etal., 2015; Janssen etal., 2006; Magliano & Millis, 2003;

Seipel et al., 2017), and/or did not distinguish between the processing of central

and peripheral text ideas (Denton etal., 2015; Janssen etal., 2006; Laing & Kamhi,

2002; Magliano & Millis, 2003).

In one recent study, Seipel etal. (2017) used a think aloud procedure to exam-

ine the manner that 3rd-to-5th grade poor and good comprehenders process impor–

tant ideas of narrative texts. Ideas were considered important if they had more than

five causal connections with other text ideas or contained a main or supporting goal.

1 3

The origin ofcentrality deficit intext memory andcomprehension…

Think-aloud responses, which were collected after each sentence of two short sto-

ries, were classified into various processing categories, including connections with

prior text ideas, elaborations by prior knowledge, predictions, monitoring responses,

paraphrases, text repetitions, irrelevant associations, and emotional responses (see

Seipel etal.’s Table2, next section, and “Appendix 1’’ for further clarification of

these categories). The first four categories were considered as high-level processes

that support the construction of a coherent, valid text representation, whereas the

last four categories were considered as low-level processes, which are less effective

in constructing useful, comprehensible representation of the text. Seipel etal. did

not find a significant difference between poor and good comprehenders in process-

ing the important ideas, although PC generated fewer high-level responses than GC

across text sentences. They did find that PC generated fewer high-level responses

than GC at the end of the text, and more low-level responses following poorly-con-

nected sentences preceding a highly-connected sentence, that completed a sub-goal

in the story. These findings suggest that PC employ similar processes as GC in pro-

cessing important ideas during reading of narrative texts. Nevertheless, a follow-up

study which will compare the likelihood of specific responses (e.g., connections)

following central (i.e., important) versus peripheral ideas by decoding-matched

groups of good and poor comprehenders, is required to draw firmer conclusions.

Furthermore, the quality of online text processing by poor and good comprehend-

ers should be compared directly with oﬄine performance in comprehension and/or

memory tasks completed by the same sample of participants.

The present study

The aim of the present study was three-fold. First, we planned to replicate the cen-

trality deficit found in PC (i.e., inferior recall of central ideas; Yeari & Lev, 2020).

This replication was essential, given that one of the two studies that compared the

size of centrality effect in GC and PC did not observe any deficit for PC (Yuill &

Joscelyne, 1988). Second, we explored the origin of the centrality deficit found in

PC following the reading of expository texts (Yeari & Lev, 2020). Specifically, a

think-aloud method was employed in this study to examine the hypothesis that poor

formation of intra-text connections during reading underlies the centrality defi-

cit found in text recall by PC after reading. Third, utilizing the diverse think-aloud

responses, we examined the overall quality of text processing adopted by PC dur–

ing reading, in order to shed more light on the cognitive underpinnings underlying

their poor comprehension and memory after reading (Cragg & Nation, 2006; Hua

& Keenan, 2014; Laing & Kamhi, 2002; Meyer etal., 1980; Yeari, Elentok etal.,

2017).

To address these goals, adolescents with poor and good comprehension, matched

on word reading ability and non-verbal intelligence, were asked to state aloud what-

ever comes to their mind during the reading of two expository texts. To encourage

natural, spontaneous text processing as much as possible, we allowed participants to

avoid responding whenever they felt that nothing came to their mind. Think-aloud

responses were classified into eight processing categories (Linderholm & Van den

M.Yeari, S.Lantin

1 3

Broek, 2002; Seipel etal., 2017) (see “Appendix 1’’ for a detailed, complete coding

schema): (1) Connective inferences—the participants explicitly reinstated text ideas

that appeared previously in the text, with or without supplementary information

inferred from prior knowledge; (2) Elaborative inferences—the participants inferred

relevant, explanatory information from prior knowledge, without reinstating prior

text ideas; (3) Predictions—the participants specified anticipation of possible infor–

mation that might appear later in the text; (4) Monitoring comments—the partici-

pants reflected on their own mental processes, such as comprehension (e.g., “I did

not understand the sentence”), attention (e.g., “I was not concentrating”), reasoning

(e.g., “it makes sense to me”), or memory processes (e.g., “I’m trying to remember

what was mentioned before”); (5) Text repetitions—the participants repeated infor–

mation that appeared in the text idea they had just read; (6) Paraphrases—the par–

ticipants used their own words to specify information which appeared in the text

idea; (7) Irrelevant associations—the participants stated information that was irrel-

evant, unnecessary, and/or did not support text comprehension; (8) Evaluative com-

ments—participants specified their own subjective opinions of the text, which were

unrelated to its specific semantic content (e.g., “the text is interesting”).1

After participants had finished the reading and thinking aloud tasks, they freely

recalled text ideas and answered comprehension questions on the texts. In addi-

tion to the think-aloud setting, the participants read one additional expository text,

sentence-by-sentence, without the requirement to think-aloud during reading. This

control condition was added to ensure that any differences expected between poor

and good comprehenders in the tasks after reading would not be an outcome of the

think-aloud procedure itself employed during reading (Laing & Kamhi, 2002; Van

Neste, Hayden, Lorch, & Milich, 2015).

To address the first research goal, i.e., replicating the centrality deficit found in

PC, we compared the extent to which GC and PC recalled central and peripheral

ideas. We expected to replicate both centrality effect and centrality deficit in PC

(Hypothesis 1). That is, we predicted that both GC and PC would recall more central

than peripheral ideas after reading (i.e., centrality effect). However, this difference

should be smaller in PC as compared to GC (i.e., centrality deficit), due to a greater

reduction in recall of central ideas as compared to peripheral ideas.

To address the second research goal, i.e., exploring the origin of centrality deficit

in PC, we compared the extent to which GC and PC established intra-text connec-

tions from and to central and peripheral ideas during reading. Connections ‘from’

referred to those text ideas which induced connective inference responses described

above, whereas connections ‘to’ refer to those text ideas which were reinstated in

the connective inference responses. This distinction was important, because it is

possible that central ideas establish more intra-text connections than peripheral

ideas throughout reading by their reinstatement during the processing of succeed-

ing ideas (i.e., connections to) more than by connecting them to preceding ideas

during their own processing (i.e., connections from). Based on the hypothesis that

poor formation of intra-text connections underlies the centrality deficit found in PC

1 Responses were also scored as valid-plausible or invalid-implausible ones. However, we analyzed only

the valid responses, because the proportions of invalid responses were low (> 3%) and similar in the two

groups.

The origin ofcentrality deficit intext memory andcomprehension…

(Yeari & Lev, 2020), we predicted that PC would establish fewer intra-text connec-

tions than GC, particularly from and/or to central ideas (Hypothesis 2). Operation-

ally, we expected that PC would exhibit a smaller difference in comparison to GC

in the number of connective inferences they would generate following the central as

compared to peripheral ideas (i.e., connections from text ideas), and/or in the num-

ber of central as compared to peripheral ideas mentioned in their connective infer–

ences (i.e., connections to text ideas).

To address the third research goal, i.e., exploring the cognitive underpinnings

underlying poor comprehension and memory of PC after reading, we compared the

overall quality of text processing employed by GC and PC during reading. Specifi-

cally, we distinguished between responses reflecting high-level, deep text processing,

such as connective inferences, elaborative inferences, predictions and monitoring,

and responses that represent low-level, surface text processing, such as text repeti-

tions, paraphrases, irrelevant associations and affective evaluations (King, 2007;

Linderholm & Van den Broek, 2002; McCrudden, Schraw, & Hartley, 2006; Press-

ley & Aﬄerbach, 1995; Seipel etal., 2017; Van den Broek, 1994; Van den Broek,

Lorch, Linderholm, & Gustafson, 2001). Then we examined the association between

online performance measured by the think-aloud task and oﬄine performance meas-

ured by the free recall and comprehension tasks. We expected that PC would exhibit

poor recall and comprehension after reading (Hypothesis 3; Cain & Oakhill, 1999;

Cragg & Nation, 2006; Hua & Keenan, 2014; Yeari, Elentok etal., 2017), employ