Connected text reading and differences in text reading fluency in adult readers - PubMed (original) (raw)
Connected text reading and differences in text reading fluency in adult readers
Sebastian Wallot et al. PLoS One. 2013.
Abstract
The process of connected text reading has received very little attention in contemporary cognitive psychology. This lack of attention is in parts due to a research tradition that emphasizes the role of basic lexical constituents, which can be studied in isolated words or sentences. However, this lack of attention is in parts also due to the lack of statistical analysis techniques, which accommodate interdependent time series. In this study, we investigate text reading performance with traditional and nonlinear analysis techniques and show how outcomes from multiple analyses can used to create a more detailed picture of the process of text reading. Specifically, we investigate reading performance of groups of literate adult readers that differ in reading fluency during a self-paced text reading task. Our results indicate that classical metrics of reading (such as word frequency) do not capture text reading very well, and that classical measures of reading fluency (such as average reading time) distinguish relatively poorly between participant groups. Nonlinear analyses of distribution tails and reading time fluctuations provide more fine-grained information about the reading process and reading fluency.
Conflict of interest statement
Competing Interests: The authors have declared that no competing interests exist.
Figures
Figure 1. Illustration of the reading task with sentence units.
Participants revealed each new text unit by pressing a space bar. Text would build up on the screen, line by line, until the whole screen was filled with text. When the whole screen was filled with text and the participant hit space once more, the screen would blank and the next text unit (word, phrase, or sentence) would appear in the upper-left corner.
Figure 2. Average reading times for the three text units (words, phrases, and sentences) by reader group and repeated reading.
The upper panel shows reading times for word (left column), phrases (middle column), and sentences (right column). As can be seen, it takes longer to read bigger text units. Also, as can be seen in the upper-right panel, sentence reading times are faster for graduate readers compared to undergraduate readers, and generally decrease with repeated reading as well. The lower panel shows the same data, when the reading times for phrases and sentences were scaled down to the number of words. Average word reading times do not differ reliably between the three text-unit conditions.
Figure 3. Time-series of average reading time for words (top row), phrases (middle row), and sentences (bottom row).
The left column presents the time series of mean reading times (standardized by the number of words for the phrase and sentence conditions). Word reading times become faster as story reading progresses, especially during the first 500–600 words (which equals the first 60–70 phrases or the first 40–50 sentences). The right column shows log-log plots of the time series of mean reading times: mean log reading times decrease linearly with log position of the word, phrase, or sentence in the story. That is, word reading times decrease as a power-law function of their position in the story. The slope of a linear regression line fitted to the log-log reading times estimates the magnitude of the decrease, which is −0.09 for words, −0.10 for phrases, and −0.09 for sentences.
Figure 4. Distribution properties of reading times.
The upper panel displays the aggregated response time distributions for the three text units, words, phrases, and sentences (in that order). Note that the x-axes for phrases and sentences reach from 0 to 20 seconds, while the x-axis for word reading times reaches from 0 to 2.5 seconds. The lower panel displays the average α exponent of the response time distributions by text unit, reader group, and repeated reading. The distribution tails for word and sentence reading times are indeed steeper for repeated reading than for reading once.
Figure 5. Modes of the reading time distributions for words, phrases, and sentences.
The modes of sentence reading times decrease for undergraduate readers, while no such change is observed for graduate readers. There were no changes in the modes of word- or phrase-unit reading times.
Figure 6. Average correlation between word frequency (top panel), word length (middle panel), and co-occurrences (lower panel) with reading times.
The correlation is plotted as a function of the number of text units (words, phrases and sentences) considered, starting with the first 10 text units. As can be seen, the correlation strength is relatively high for the very first bins for each text unit, but dropping off as more and more words/phrases/sentences are considered.
Figure 7. Recurrence analysis of ‘the rain in Spain stays mainly in the plain’.
Left: Recurrence structure obtained from ‘the rain in Spain’ with delay of 1 and dimensionality of 3 (i.e., the text vector is shifted by 1 symbol 2 times and identical recurring trigrams are plotted as dots in the recurrence plot. Right: The recurrence structure of ‘the rain in Spain stays mainly in the plain’ is reduced to a single vector of redundancies, where the number of dots in each column are summed up for the lower-right half of the plot (the plot is symmetrical above and below the diagonal).
Figure 8. Results of recurrence analysis for words (left column), phrases (center column), and sentences (right column).
The top row displays the values for %REC, the middle row displays the values for %DET, and the bottom row displays the values for %LAM. %REC and %LAM dropped with repeated reading in the word-unit condition, while no effects were apparent in the phrase and sentence reading conditions. However, %REC, %DET, and %LAM were lower for phrase and sentence-unit reading compared to word-unit reading.
Figure 9. Results of cross recurrence analysis for words (left column), phrases (center column), and sentences (right column).
The top row displays the values for shared %REC, the middle row displays the values for shared %DET, and the bottom row displays the values for shared %LAM. The three text-unit conditions differ in the effects on the recurrent measures: In the word-unit condition, shared %REC, %DET, and %LAM decrease with repeated reading for both reader groups. In the phrase unit condition, shared %DET and %LAM increase with repeated reading for graduate readers, but decrease (or do not change) for undergraduate readers with repeated reading. In the sentence-unit condition, shared %REC increases for both groups with repeated reading, while shared %DET and %LAM only increases for undergraduate readers with repeated reading.
Figure 10. Monofractal α values for word (left), phrase (center), and sentence (right) reading times.
Word reading times become increasingly long-range correlated with repeated reading, while no such change is observed for phrase and sentence reading times. While word reading times show generally a high degree of long-range correlation, phrase and sentence reading time fluctuations are much more locally determined.
Figure 11. Multifractal spectrum width for word (left), phrase (center), and sentence (right) reading times.
Multifractality decreases with repeated reading in the word-unit condition, while multifractality is absent in the phrase and sentence reading conditions.
References
- Rawlinson GE (1976) The significance of letter position in word recognition. PhD Thesis. Nottingham: Psychology Department, University of Nottingham.
- Epelboim J, Booth JR, Ashkenazy R, Taleghani A, Steinman RM (1997) Fillers and spaces in text: The importance of word recognition during reading. Vision Res 37: 2899–2914. -PubMed
- Luce PA, Cluff MS (1998) Delayed commitment in spoken word recognition: evidence from cross-modal priming. Percept Psychophys 60: 484–490. -PubMed
- Rayner K, Pollatsek A (1994) The psychology of reading. Englewood Cliffs, NJ: Prentice Hall.
- Clifton C Jr, Duffy SA (2001) Sentence and text comprehension: Roles of linguistic structure. Annu Rev Psychol 52: 167–196. -PubMed
Publication types
MeSH terms
LinkOut - more resources
Full Text Sources
Other Literature Sources