1
Journal of Neuropsychology (2015) © 2015 The British Psychological Society www.wileyonlinelibrary.com
Word recognition in Alzheimer’s disease: Effects of semantic degeneration Fernando Cuetos1, Noemı Arce1, Carmen Martınez2 and Andrew W. Ellis3,* 1
University of Oviedo, Spain Cabue~ nes Hospital, Gij on, Spain 3 University of York, UK 2
Impairments of word recognition in Alzheimer’s disease (AD) have been less widely investigated than impairments affecting word retrieval and production. In particular, we know little about what makes individual words easier or harder for patients with AD to recognize. We used a lexical selection task in which participants were shown sets of four items, each set consisting of one word and three non-words. The task was simply to point to the word on each trial. Forty patients with mild-to-moderate AD were significantly impaired on this task relative to matched controls who made very few errors. The number of patients with AD able to recognize each word correctly was predicted by the frequency, age of acquisition, and imageability of the words, but not by their length or number of orthographic neighbours. Patient Mini-Mental State Examination and phonological fluency scores also predicted the number of words recognized. We propose that progressive degradation of central semantic representations in AD differentially affects the ability to recognize low-imageability, low-frequency, lateacquired words, with the same factors affecting word recognition as affecting word retrieval.
Two of the cognitive problems most commonly reported in the early stages of Alzheimer’s disease (AD) are difficulty remembering recent events (episodic memory) and difficulty understanding and producing words (semantic memory; see Altmann & McClung, 2008; Lambon Ralph et al., 2001; Taler & Phillips, 2008; Verma & Howard, 2012). Problems with word retrieval have been studied intensively through the use of picture naming and other tasks (Chertkow & Bub, 1990; Chertkow, Bub, & Caplan, 1992; Cuetos, Martinez, Martinez, Izura, & Ellis, 2003; Hodges, Salmon, & Butters, 1991, 1992; Martin & Fedio, 1983). Problems with word recognition and comprehension in AD have also been documented and, like the problems in word retrieval, have mainly been attributed to impairments affecting the mental representations of concepts, including the semantic representations of familiar objects. Deficits in word recognition remain, however, underinvestigated compared with deficits in word retrieval and production. Evidence for a semantic basis to the lexical problems in AD includes the following: (1) the ability of patients to name specific objects is predicted by the amount of semantic
*Correspondence should be addressed to Andrew W. Ellis, Department of Psychology, University of York, York YO10 5DD, UK (email: [email protected]). DOI:10.1111/jnp.12077
2
Fernando Cuetos et al.
information they retain about those objects (Garrard, Lambon Ralph, Patterson, Pratt, & Hodges, 2005; Lambon Ralph, Patterson, & Hodges, 1997); (2) semantic naming errors (such as naming penguin as duck, or cow as milk) predominate in the early stages of the disease (Barbarotto, Capitani, Jori, Laiacona, & Molinari, 1998; Cuetos, Gonzalez-Nosti, & Martınez, 2005; Cuetos, Rodrıguez-Ferreiro, Sage, & Ellis, 2012; Hodges et al., 1991; Huff, Corkin, & Growden, 1986; Martin & Fedio, 1983; Rodrıguez-Ferreiro, Davies, GonzalezNosti, Barb on, & Cuetos, 2009); and (3) distinctive semantic information about concepts is lost sooner than information that is shared with many other concepts (Flanagan, Copland, Chenery, Byrne, & Angwin, 2013) with the result that patients can generate fewer atypical members of particular categories than typical members (Perri, Zannino, Caltagirone, & Carlesimo, 2012; Sailor, Antoine, Diaz, Kuslansky, & Kluger, 2004). One of the issues addressed in this paper is the extent to which the same properties of words predict success or failure in word retrieval and word recognition by patients with AD. Two factors that have emerged as consistent predictors of object naming accuracy in AD are word frequency and age of acquisition (AoA), with patients being better at retrieving and producing high-frequency names and names learned early in life than lowfrequency names and names learned later in life (Cuetos, Rodrıguez-Ferreiro et al., 2012; Rodrıguez-Ferreiro et al., 2009; Tippett, Meier, Blackwood, & Diaz-Asper, 2007). Theoretical interpretations of those effects have proposed that the semantic representations of concepts learned early in life and activated with high frequency thereafter are richer and/or easier to access and that the semantic representations of concepts are learned later in life and activated less often. Richer, more accessible semantic representations survive the early stages of AD better than poorer, less accessible representations and are, for example, able to drive successful name retrieval in patients in the mild-to-moderate stages of AD (see Ellis, 2011; for a review). The apparent lack of influence of the complexity of object pictures or the length of object names on picture naming in patients with AD has been interpreted as indicating that the primary impairment is semantic rather than, for example, perceptual or phonological (Albanese, 2007; Cuetos et al., 2005; Cuetos, Rodrıguez-Ferreiro et al., 2012; Rodrıguez-Ferreiro et al., 2009; Silveri, Cappa, Mariotti, & Puopolo, 2002; Tippett et al., 2007). One problem with investigating word recognition in AD and other neuropsychological conditions is that the tasks used to assess recognition and comprehension often involve presenting more than one word or picture at a time; for example, asking patients to point to the picture that matches a target word from a set of alternatives, or indicate whether pairs of words have similar or dissimilar meanings. When multiple words or pictures are presented, it becomes difficult to separate the properties of the target words from the properties of the other words or pictures that accompany them. The lexical decision task has sometimes been employed in an effort to overcome these problems. In lexical decision, participants are shown a sequence of words interleaved with invented nonwords (i.e., strings of letters that look like words but happen not to be; for example, flupe or quentole). In the traditional form of the lexical decision task, one word or non-word is presented at a time and the participant is required to indicate whether each item is a word or not. There is no requirement to demonstrate any understanding of the meanings of the words presented. Using visual presentation of written words and non-words, Chertkow and Bub (1989, 1990), Chertkow et al. (1992), Cuetos et al. (2003) and Du~ nabeitia, Marın, and Carreiras (2009) reported that patients with AD showed similar levels of accuracy in lexical decision as healthy, matched controls. In contrast, Madden, WelshBohmer, and Tupler (1999) reported less accurate lexical decision in patients with AD
Word recognition in Alzheimer’s disease
3
using written presentation while Glosser, Kohn, Friedman, Sands, and Grugan (1997) reported less accurate performance when the stimuli were spoken words and non-words. We will argue here that patients with AD are impaired at lexical decision relative to controls because they fail to recognize some words with which they were once familiar (and which matched controls continue to recognize). We believe that there are at least three reasons why previous studies have sometimes failed to observe reduced levels of accuracy in patients with AD. First, some studies have used object names as the words in their lexical decision tasks (Chertkow & Bub, 1989; Chertkow et al., 1992; Cuetos et al., 2003). By definition, object names are high in imageability. They tend also to be learned relatively early in life. If word recognition in patients with AD is better for highimageability, early-acquired words than for low-imageability, late-acquired words, then object names may be relatively easy for patients with AD to recognize and therefore not particularly well suited to identifying problems with word recognition. A second issue was highlighted by Madden et al. (1999) who found a decrease in lexical decision accuracy in patients with AD relative to controls that was greater for nonwords than for words. That is, patients appeared able to recognize most of the words correctly while miscategorising many non-words as words. This may indicate that when patients with AD are unsure as to whether an item is a word or a non-word, they tend to categorize it as a word. A Yes bias of this sort will sustain a spuriously high level of apparently correct responses to words and will only be revealed in the standard lexical decision task if overall accuracy among the patients with AD is not at ceiling and if error rates to non-words are also analysed. Madden et al. (1999) addressed this issue using a signal detection analysis which demonstrated a substantial bias towards responding Yes to non-words, but also an impaired sensitivity to words over and above the change in response bias. A simple way to eliminate response bias is to modify the lexical decision task to a version in which participants are presented on each trial with (for example) one word and one non-word and are asked to indicate which item is the real word. With a requirement to pick out the word on every trial, response biases are removed. Baddeley, Emslie, and Nimmo-Smith’s (1993) ‘Spot the Word’ test involved just such a two-alternative forcedchoice version of lexical decision (Baddeley & Crawford, 2012). Using that test, Law and O’Carroll (1998) found a trend towards more errors in patients with AD compared with controls, while Beardsall and Huppert (1997) reported more errors in patients with mildto-moderate AD than in patients with minimal dementia. But although these studies dealt with the issue of response bias, they illustrate a third potential problem. If lexical selection or lexical decision tasks use half words and half non-words, participants have a 50% chance of making a correct response on each trial even if they actually have no idea which of the stimuli are words and which are non-words. A 50% chance rate makes accuracy scores relatively insensitive to impairments. The lexical selection task can be made more sensitive by increasing the number of non-words that accompany each word target, thereby reducing the probability of selecting the real word purely by chance. Cuetos, Herrera, and Ellis (2010) presented three non-words with each real word in a lexical selection task, thereby reducing the chance of a fortuitously correct response to 25%. Patients with mild-to-moderate AD correctly selected the real word on an average of 88% of trials, while matched controls selected the word on an average of 99% of trials. Although well above chance, patients with AD performed significantly worse than controls. If the problems affecting word recognition in AD share a common origin with the problems affecting word retrieval (e.g., impairment to semantic representations affecting both word recognition and production), we would expect the same factors to influence
4
Fernando Cuetos et al.
lexical selection as have been shown to influence word retrieval. As noted above, one of the factors consistently found to influence word retrieval by patients with AD in naming tasks is AoA: All other things being equal, patients are better able to name objects with early- than late-acquired names (Albanese, 2007; Catling, South, & Dent, 2013; Cuetos, Rodrıguez-Ferreiro et al., 2012; Holmes, Fitch, & Ellis, 2006; Marquez, Cappa, & Sartori, 2011; Rodrıguez-Ferreiro et al., 2009; Silveri et al., 2002; Tippett et al., 2007). Effects of AoA have also been observed when patients are asked to generate words belonging to different semantic categories (Forbes-McKay, Ellis, Shanks, & Venneri, 2005; Sailor, Zimmerman, & Sanders, 2011). In addition, Perri et al. (2012) found that patients with mild AD were able to generate more semantic features for early- than late-acquired concepts. Cuetos, Rodrıguez-Ferreiro et al. (2012) found worse lexical selection performance for late- than early-acquired words in the lexical selection task using sets of words matched on frequency, imageability, and other factors. AoA would therefore appear to be one property of words and concepts that affects the accuracy of both word retrieval and word recognition in patients with AD. A number of recent studies have reported effects of word frequency on object naming in patients with AD that were independent of the effects of AoA and other factors (Cuetos, Rodrıguez-Ferreiro et al., 2012; Rodrıguez-Ferreiro et al., 2009; Tippett et al., 2007). We therefore predicted an effect of word frequency on lexical selection in patients with AD independent of the effects of any other factors. It is hard to evaluate the effects of imageability on object naming because objects tend to score highly on imageability, so the range of values available for manipulation is limited. Failures to find the effects of imageability on naming accuracy in patients with AD (Albanese, 2007; Cuetos et al., 2005; Cuetos, Rodrıguez-Ferreiro et al., 2012; Rodrıguez-Ferreiro et al., 2009) cannot form the basis of firm predictions regarding possible effects in lexical selection where a much wider range of imageability values can be employed. Two studies have, however, reported effects of imageability/concreteness in patients with AD using alternative tasks. Rissenberg and Glanzer (1987) reported that the impairment shown by patients with AD at producing words in response to definitions was substantially greater when the target words were abstract than when they were concrete while Peters, Majerus, De Baerdemaeker, Salmon, and Collette (2009) found better immediate serial recall of highthan low-imageability words in patients with AD. Effects of imageability (or concreteness) are generally regarded as indicators of semantic involvement in particular tasks, the assumption being made that the limited perceptual experience associated with abstract words causes their semantic representations to be less detailed than the representations of concrete words (Hoffman, Jones, & Lambon Ralph, 2013). If the underlying deficit in patients with AD is primarily semantic, and if imageability is a semantic variable, then we would predict an effect of imageability on lexical selection in the present study. Some words in the language resemble many other words, while others are more distinctive in their appearance. One measure of distinctiveness is orthographic neighbourhood size (N), defined as the number of other words that differ from a target word by a single letter (Coltheart, Davelaar, Jonasson, & Besner, 1977). Distinctive words have low N values, while typical words have higher values. The only study we are aware of that has investigated the effects of N on object naming in patients with AD is by RodrıguezFerreiro et al. (2009) who found no significant effect. Du~ nabeitia et al. (2009) compared lexical decision accuracy and reaction times to words with many or few orthographic neighbours in patients with AD and in controls. Both groups made more errors to low than high N words, but the difference in favour of the high N words was smaller in the patients with AD and not significant. That is, in terms of its effect on accuracy, the effect of N was
Word recognition in Alzheimer’s disease
5
reduced in patients with AD compared with controls. Du~ nabeitia et al. (2009) also reported comparable effects of N on reaction times in patients with AD and in controls. Accuracy of responses to non-words was not reported, so the possibility of a Yes bias in the patients cannot be excluded. Finally, studies of word retrieval in AD have consistently failed to find effects of word length. When word frequency and other factors are taken into account, patients are as likely to be able to retrieve and produce a long object name correctly as a short name (Cuetos, Rodrıguez-Ferreiro et al., 2012; Rodrıguez-Ferreiro et al., 2009; Silveri et al., 2002; Tippett et al., 2007). We are not aware of any previous investigations of the influence of length on word recognition in patients with AD, but if the principal deficit in AD is semantic rather than perceptual, orthographic, or phonological, there is no reason to expect an effect of length on lexical selection performance in patients with AD. The stimuli of interest in the present study were 150 words that varied on word frequency, AoA, imageability, N, and length. These were chosen to be words that the controls were expected to know. On each trial of the lexical selection task, one written word was accompanied by three non-words that were matched to the word on number of letters and syllables. Participants were simply asked to indicate which item in the array was a real word. The participants were 40 patients with mild-to-moderate AD and 25 matched controls. We predicted that the patients with AD would have difficulty recognizing the more low-frequency, late-acquired, and low-imageability words. We did not expect to find an effect of length. The literature was ambiguous as to whether an effect of N might be observed. In presenting the results, we report one set of correlations which show the relationships across patients between age, years of education, Mini-Mental State Examination (MMSE) scores, scores on two fluency tasks, and lexical selection accuracy. We then present an item analysis showing the correlations across words between frequency, AoA, imageability, etc. and the accuracy with which the patients with AD as a group responded to those words. The main statistical analysis is, however, a mixed-effects multiple regression. Where a conventional multiple regression would take a mean level of accuracy for each word across patients and use it to look at the effects of word properties like frequency and length, a mixed-effects multiple regression takes every response from every participant to every word (i.e., 40 patients with AD 9 150 words = 6,000 data points) and allows the researcher to analyse the effects of participant and stimulus characteristics in the same analysis which is more sensitive to effects than a traditional multiple regression (cf. Cuetos, Rodrıguez-Ferreiro et al., 2012; Gonzalez-Nosti, Barb on, Rodrıguez-Ferreiro, & Cuetos, 2013). Our mixed-effects analysis combined participant MMSE and fluency scores with frequency, AoA, and the other lexical variables.
Methods Participants Forty patients with probable dementia of Alzheimer’s type (35 female and 5 male aged 66– 91 years) took part in the study. The patients were selected on the basis of their medical history, information from a knowledgeable informant, a CT or MRI scan, and a neuropsychological evaluation which included the MMSE (Folstein, Folstein, & McHugh, 1975) along with tests of semantic fluency (naming as many animals as possible in 1 min) and phonological fluency (producing as many words beginning with ‘s’ as possible in 1 min). The diagnosis of probable AD was made according to the Neurological and
6
Fernando Cuetos et al.
Communicative Disorders and Stroke – Alzheimer’s Disease and Related Disorders (NINCDS-ARDA criteria; McKhann et al., 1984; revised by McKhann et al., 2011). Patients also met the DSM-IV criteria for dementia of the Alzheimer’s type (American Psychiatric Association, 2000). The study was approved by the Ethics Committee of the Hospital Cabue~ nes, Gij on. Informed consent was obtained from all participants and from patients’ caregivers prior to the study. The controls were 25 healthy adult volunteers (21 female and 4 male) matched to the patients on age and years of education (Table 1). None had a psychiatric history, sensory deficiencies, or medical conditions that could impair performance on the neuropsychological tests. Patients with AD differed significantly from the controls on MMSE score, t (63) = 10.62, p < .001, semantic fluency, t(63) = 8.77, p < .001, and phonological fluency, t(63) = 6.45, p < .001, but not on age, t(63) = 1.10, p = .276, or years of education, t(63) = 0.07, p = .946.
Materials One hundred and fifty Spanish words were selected for this study covering a range of values for AoA, word frequency, imageability, length, and N. The words were taken from the ratings study by Cuetos, Samartino, and Ellis (2012) and were items that a high proportion of adults in that study (30 healthy adults aged 61–85 years) recognized as familiar. AoA ratings were taken from the Cuetos, Samartino et al.’s (2012) study where they were obtained using a scale from 1 = learned before the age of 2 years to 8 = learned after the age of 20. Such ratings have been shown to correlate highly with objective measures of AoA (Brysbaert & Ellis, in press; Ellis, 2011). Word frequency values were taken from the Subtlex-Esp database (Cuetos, Gonzalez-Nosti, Barb on, & Brysbaert, 2011) which is based on the frequencies with which words occur in a corpus of 41.5 million words taken from contemporary film subtitles. Subtitle frequencies have been found to predict performance in word recognition experiments better than frequency counts based solely on written texts (Brysbaert & New, 2009). The words had a mean of 14.45 occurrences per million words of Spanish (range 0.19–311.85). Imageability values were taken from the LEXESP database (Sebastian, Martı, Carreiras, & Cuetos, 2000) where adult participants rated words on a 7-point scale from 1 = very hard to conjure up a mental image to 7 = very easy to conjure up a mental image. The words had a mean imageability value of 5.00 (range 2.07–6.67). Values of N were taken from Perez, Alameida, and Cuetos (2003). The words had a mean of 2.11 neighbours (range 0–17). Finally, word length was measured as the number of letters in each word Table 1. Summary of participant characteristics and scores on the MMSE, semantic fluency, and phonological fluency tasks Age Patients with AD Mean 78.70 SD 5.98 Controls Mean 77.20 SD 4.13
Years of education
MMSE
Semantic fluency
Phonological fluency
8.10 3.20
19.88 4.23
7.15 3.44
5.23 3.57
8.16 3.83
29.08 1.15
15.88 4.56
10.96 3.32
Note. AD = Alzheimer’s disease, MMSE = Mini-Mental State Examination, SD = standard deviation.
Word recognition in Alzheimer’s disease
7
(mean 6.15; range 4–9). During item selection, an effort was made to reduce the intercorrelations between the variables while maintaining a reasonable spread of values on each variable. Four hundred and fifty non-words were generated, three for each target word. The non-words were pronounceable, word-like letter strings which could be words in Spanish but happen not to be (e.g., lefa, milusa, and pasorento). The three non-words that accompanied each target word were matched to the target on the number of letters and syllables. To check that there were no confounds between the properties of the target words and the characteristics of the non-words that accompanied them, we calculated the mean length, N, and bigram frequencies of the non-words used on each trial. The correlations between the mean length, N, and bigram frequencies of the non-words and the frequency, AoA, and imageability of the accompanying words were all non-significant. It is unlikely, therefore, that any effect of frequency, AoA, or imageability found for the recognition of real words was an indirect consequence of differences in the ease or difficulty of rejecting the non-words that were presented with the words on each trial. The three non-words and one word were presented on the screen of a laptop in a square formation (two above and two below) using black, lower case font on a white background. Words were distributed evenly across the four possible positions, with a real word occurring at each position either 37 or 38 times.
Procedure Participants were tested individually in a quiet room in the Neurology Unit of the Cabue~ nes Hospital, Gij on. The experimenter first explained the lexical selection task using several examples. When the experimenter was satisfied that the participant understood the task, the experimental stimuli were presented one set at a time. On each trial, participants were asked to look at the four alternatives and indicate which was the real word by pointing to it. Occasionally, participants recognized the real word quickly and named it rather than pointing to it. Those responses were accepted, with the patient being reminded of the requirement to point to the real words rather than name them. To minimize non-verbal cues, the experimenter stood to the side and slightly behind the patient, noting the response on each trial by marking the chosen item on a response sheet which showed the four alternatives. No feedback was given during the experiment. The MMSE and fluency tasks were administered on the same day as the lexical selection task.
Results As expected, performance of the controls on the lexical selection task was close to ceiling. Controls identified a mean of 147.4 of the 150 words correctly (98.3%), with 20 of the 25 controls responding correctly to 147 or more of the words. The patients with AD identified a mean of 123.2 words correctly (82.1%). That is well above the chance rate of 25% but significantly worse than the accuracy of the controls, t(63) = 6.46, p < .001. Having established that the patients with AD were impaired on the lexical selection task, further analyses focused on the patient data. Table 2 shows the correlations between age, years of education, MMSE, semantic fluency, phonological fluency, and lexical selection for the patients with AD. Lexical selection scores correlated significantly with number of years of education, MMSE, semantic fluency, and phonological fluency.
8
Fernando Cuetos et al.
Table 2. Correlations between age, years of education, MMSE, semantic fluency, phonological fluency, and lexical selection for the patients with AD
Age Years of education MMSE Semantic fluency Phonological fluency Lexical selection
Age
Years of Educ.
MMSE
Sem. fluency
–
.307 –
.064 .156 –
.015 .082 .589*** –
Phon. fluency .048 .107 .410** .605*** –
Lexical selection .036 .377* .688*** .433** .455** –
Notes. Educ. = education, MMSE = Mini-Mental State Examination, Sem. = Semantic, Phon. = Phonological. *p < .05; **p < .01; ***p < .001.
Table 3 shows the correlations across items between the word properties (frequency, AoA, etc.) and the number of patients with AD who selected each word correctly. Lexical selection scores for the 150 words correlated significantly with word frequency, AoA, and imageability, but not with N or length. When the predictor variables are correlated with each other as well as with the dependent variable (number of patients selecting each word correctly), some form of regression analysis is required to determine which predictor variables are exerting genuinely independent effects on the dependent variable. We employed a mixed-effects multiple regression model implemented through the lme4 package in R (R Development Core Team, 2012, version 2.15). The model had patients and items as random intercepts and the predictors as fixed effect factors or covariates (cf. Baayen, Davidson, & Bates, 2008; Kuperman, Schreuder, Bertram, & Baayen, 2009). The patient characteristics employed as predictors were MMSE scores and scores on the semantic and phonological fluency tasks. The lexical predictors were word frequency, AoA, imageability, length, and N. Log values of the predictors were used to reduce skew. Collinearity was assessed by calculating the variance inflation factor (VIF) for each predictor. VIF provides a measure of how much larger the variance of a particular coefficient is than it would have been if that predictor was completely uncorrelated with the other predictors, with VIF values >4
Table 3. Correlations among the predictor variables and between the predictor variables and the number of patients with AD who selected each word correctly in the lexical selection test AoA AoA Frequency Imageability Length N Lexical selection
–
Freq
Imag
.281*** –
.531*** .029 –
Length .143 .029 .152 –
N .244** .099 .121 .648*** –
Lexical selection .431*** .234** .344*** .057 .134 –
Notes. AoA = age of acquisition, Freq = Frequency, Imag = Imageability, N = Number of orthographic neighbours. *p < .05; **p < .01; ***p < .001.
Word recognition in Alzheimer’s disease
9
generally taken to indicate potential problems with multicollinearity (Cohen, Cohen, West, & Aiken, 2003). The VIF values for the predictors in this study ranged from 1.275 to 1.714, indicating that multicollinearity was not a problem and that the analysis was valid. The analysis found significant effects of MMSE, Z = 4.55, p < .001, and phonological fluency, Z = 2.25, p < .05, among the participant variables and word frequency, Z = 3.28, p < .001, AoA, Z = 6.21, p < 001, and imageability, Z = 4.26, p < .001, among the lexical variables. In the context of these other predictors, the effects of semantic fluency score, Z = 0.52, p = .600, neighbourhood size (N), Z = 0.39, p = .692, and word length, Z = 0.56, p = .574, did not approach significance.
Discussion The words employed in the present study were chosen to be ones that the healthy controls were expected to recognize (and therefore, by inference, words that the patients with AD would have been able to recognize before the onset of their illness). As expected, control performance was at or close to ceiling while the patients with AD identified significantly fewer real words than the controls. We note that the patients with AD in our study were predominantly female. That would be expected from the longer life expectancy of women, but a meta-analysis by Irvine, Laws, Gale, and Kondel (2012) found that cognitive functions are more severely and more widely affected in women with AD than in men, even allowing for differences in age, education, or dementia severity. The controls in the present study had the same proportion of men and women as the patients with AD. The lexical selection task eliminates the problem of response bias that can affect traditional lexical decision and reduces the probability of making a correct response by chance. The finding that lexical selection was impaired in the patients with AD supports previous demonstrations of impaired lexical decision/selection in AD (Beardsall & Huppert, 1997; Cuetos et al., 2010; Glosser et al., 1997; Madden et al., 1999) and suggests that reports of near-normal lexical decision accuracy in patients with AD may have been due to a combination of response biases, high chance rates, and the use of relatively easy, high-imageability words. Having established that the patients with AD failed to recognize words they would previously have known, further analyses focused on the patient data. The mixed-effects regression analysis found that lexical selection accuracy was predicted by the MMSE and phonological fluency scores of the patients and by the frequency, AoA, and imageability of the words. In the context of those predictors, the effects of semantic fluency scores and the length and N values of words were not significant. Worse performance at word recognition in more severe patients with lower MMSE scores is to be expected (Cuetos et al., 2010). The effect of AoA, with better recognition of early- than late-acquired words, replicates the finding of Cuetos et al. (2010). The demonstration of an additional and independent effect of word frequency is new in this context, but independent effects of frequency and AoA have been observed in studies of object naming by patients with AD (Cuetos, Samartino et al., 2012; Rodrıguez-Ferreiro et al., 2009; Tippett et al., 2007). The observed effect of imageability is in line with reported effects of that variable in other tasks (Peters et al., 2009; Rissenberg & Glanzer, 1987). Taken together, the effects of frequency, AoA, and imageability are consistent with the hypothesis that a central semantic impairment underlies much of the difficulty that patients have in both word recognition and production.
10
Fernando Cuetos et al.
We chose words for our lexical selection task that we expected healthy controls to know because we see little point in demonstrating that patients with AD are unable to recognize little-known words that controls are also unable to recognize. A connection might nevertheless be made between impaired word recognition in patients with AD and normal word recognition in healthy adults. The connection is that the same factors that cause patients with AD to be unable to recognize once-familiar words also appear to affect the speed with which healthy adults can recognize words. Effects of word frequency, AoA, and imageability have been reported on the speed with which healthy adults respond correctly to words in lexical decision (Cortese & Khanna, 2007; Cortese & Schock, 2013; Gonzalez-Nosti et al., 2013). This suggests a general rule whereby words that healthy adults recognize correctly but slowly are the words that patients with AD fail to recognize at all. We propose that the mechanism underlying these effects is that impoverished semantic representations in healthy adults mean that low-frequency, lateacquired, and low-imageability words are recognized relatively slowly in lexical decision. Those impoverished representations are more vulnerable to the degenerative effects of AD than are the richer representations of high-frequency, early-acquired, and highimageability words. Words whose semantic representations have degenerated beyond a certain point no longer seem familiar and cannot be easily distinguished from non-words in the lexical selection task (Ellis, 2011). We found no effect of orthographic neighbourhood size on lexical selection accuracy in our AD group. Studies of object naming in patients with AD have also failed to find the effects of N (Cuetos, Samartino et al., 2012; Rodrıguez-Ferreiro et al., 2009; Tippett et al., 2007). In contrast, Du~ nabeitia et al. (2009) found the effects of N on lexical decision speed and accuracy both in patients with AD and in controls. Findings from other studies have been mixed with respect to the effects of N on lexical decision speed in healthy adults (Cortese & Khanna, 2007; Cortese & Schock, 2013; GonzalezNosti et al., 2013). Balota, Cortese, Sergent-Marshall, Spieler, and Yapp (2004) found no effects of N on lexical decision speed or accuracy in young adults but effects on both speed and accuracy in older adults. This is an issue where further research is needed, including exploring different measures of the similarity between word forms. Despite this apparent anomaly, the general rule that slow recognition in healthy adults converts into recognition failure in patients with AD is, we believe, supported by the available evidence. In addition to finding no significant effects of word length and N on lexical selection accuracy in our patient group, we found no effect of semantic fluency scores. Studies have indicated that the probability that patients with AD will retrieve particular words in the semantic fluency task is influenced by both the frequency and AoA of those words (Binetti et al., 1995; Forbes-McKay et al., 2005; Marczinski & Kertesz, 2006; Sailor et al., 2011). It may be that when frequency and AoA are included as predictors of word recognition accuracy along with MMSE and phonological fluency, little or no variance remains for semantic fluency to account for.
Acknowledgements This research was supported by grant PSI2012-31913 from the Spanish Government.
Word recognition in Alzheimer’s disease
11
References Albanese, E. (2007). The “hidden” semantic category dissociation in mild–moderate Alzheimer’s disease patients. Neuropsychologia, 45, 639–643. doi:10.1016/j.neuropsychologia.2006.07.018 Altmann, L. J. P., & McClung, J. S. (2008). Effects of semantic impairment on language use in Alzheimer’s disease. Seminars in Speech and Language, 29, 18–31. doi:10.1055/s-20081061622 American Psychiatric Association (2000). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author. Baayen, R. H., Davidson, D. J., & Bates, D. M. (2008). Mixed-effects modeling with crossed random effects for subjects and items. Journal of Memory and Language, 59, 390–412. doi:10.1016/ jml.2007.12.005 Baddeley, A. D., & Crawford, J. (2012). Spot the word (2nd ed.). Oxford, UK: Pearson. Baddeley, A. D., Emslie, H., & Nimmo-Smith, I. N. (1993). The Spot-the-Word test: A robust estimate of verbal intelligence based on lexical decision. British Journal of Clinical Psychology, 32, 55– 66. doi:10.1111/j.2044-8260.1993.tb01027.x Balota, D. A., Cortese, M. J., Sergent-Marshall, S. D., Spieler, D. H., & Yapp, M. J. (2004). Visual word recognition for single-syllable words. Journal of Experimental Psychology: General, 133, 283– 316. doi:10.1037/0096-3445.133.2.283 Barbarotto, R., Capitani, E., Jori, T., Laiacona, M., & Molinari, S. (1998). Picture naming and progression of Alzheimer’s disease: An analysis of error types. Neuropsychologia, 36, 397–405. doi:10.1016/S0028-3932(97)00124-3 Beardsall, L., & Huppert, F. A. (1997). Short NART, CCRT, and Spot-the-Word: Comparisons in older and demented persons. British Journal of Clinical Psychology, 36, 619–622. doi:10.1111/ j.2044-8260.1997.tb01266.x Binetti, G., Magni, E., Cappa, S. F., Padovani, A., Bianchetti, A., & Trabucchi, M. (1995). Semantic memory in Alzheimer’s disease: An analysis of category fluency. Journal of Clinical and Experimental Neuropsychology, 17, 82–89. doi:10.1080/13803399508406584 Brysbaert, M., & Ellis, A. W. (in press). Aphasia and age of acquisition: Are early-learned words more resilient? Aphasiology. Brysbaert, M., & New, B. (2009). Moving beyond Kucera and Francis: A critical evaluation of current word frequency norms and the introduction of a new and improved word frequency measure for American English. Behavior Research Methods, 41, 977–990. doi:10.3758/BRM.41.4.977 Catling, J., South, F., & Dent, K. (2013). The effect of age of acquisition on older individuals with and without cognitive impairments. Quarterly Journal of Experimental Psychology, 66, 1963– 1973. doi:10.1080/17470218.2013.771689 Chertkow, H., & Bub, D. (1989). Priming and semantic memory loss Alzheimer’s disease. Brain and Language, 36, 420–446. doi:10.1016/0093-934X(89)90078-3 Chertkow, H., & Bub, D. (1990). Semantic memory loss in dementia of Alzheimer’s type. Brain, 113, 397–417. doi:10.1093/brain/113.2.397 Chertkow, H., Bub, D., & Caplan, D. (1992). Constraining theories of semantic memory processing: Evidence from dementia. Cognitive Neuropsychology, 9, 327–365. doi:10.1080/ 02643299208252064 Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: Lawrence Erlbaum. Coltheart, M., Davelaar, E., Jonasson, J. T., & Besner, D. (1977). Access to the internal lexicon. In S. Dornic (Ed.), Attention and performance, VI (pp. 535–555). New York, NY: Academic Press. Cortese, M. J., & Khanna, M. M. (2007). Age of acquisition predicts naming and lexical-decision performance above and beyond 22 other predictor variables: An analysis of 2,342 words. Quarterly Journal of Experimental Psychology, 60, 1072–1082. doi:10.1080/1747021070 1315467
12
Fernando Cuetos et al.
Cortese, M. J., & Schock, J. (2013). Imageability and age of acquisition effects in disyllabic word recognition. Quarterly Journal of Experimental Psychology, 66, 946–972. doi:10.1080/ 17470218.2012.722660 Cuetos, F., Gonzalez-Nosti, M., Barb on, A., & Brysbaert, M. (2011). SUBTLEX-ESP: Spanish word frequencies based on film subtitles. Psicol ogica, 32, 133–143. Cuetos, F., Gonzalez-Nosti, M., & Martınez, C. (2005). The picture-naming task in the analysis of cognitive deterioration in Alzheimer’s disease. Aphasiology, 19, 545–557. doi:10.1080/ 02687030544000010 Cuetos, F., Herrera, E., & Ellis, A. W. (2010). Impaired word recognition in Alzheimer’s disease: The role of age of acquisition. Neuropsychologia, 48, 3329–3334. doi:10.1016/j.neuropsychologia.2012.07.017 Cuetos, F., Martinez, T., Martinez, C., Izura, C., & Ellis, A. W. (2003). Lexical processing in Spanish patients with probable Alzheimer’s disease. Cognitive Brain Research, 17, 549–561. doi:10.1016/S0926=6410(03)00169-1 Cuetos, F., Rodrıguez-Ferreiro, J., Sage, K., & Ellis, A. W. (2012). A fresh look at the predictors of naming accuracy and errors in Alzheimer’s disease. Journal of Neuropsychology, 6, 242–256. doi:10.1111/j.1748-6653.2011.02025.x Cuetos, F., Samartino, T., & Ellis, A. W. (2012). Age acquisition norms from elderly Spanish people: Characteristics and the prediction of word recognition performance in Alzheimer’s disease. Psicol ogica, 33, 59–76. Du~ nabeitia, J. A., Marın, A., & Carreiras, M. (2009). Associative and orthographic neighborhood density effects in normal aging and Alzheimer’s disease. Neuropsychology, 23, 759–764. doi:10.1037/a0016616 Ellis, A. W. (2011). The acquisition, retention and loss of vocabulary in aphasia, dementia and other neuropsychological conditions. In M. Faust (Ed.), Handbook of the neuropsychology of language (Vol. 2, pp. 637–660). Oxford, UK: Blackwells. Flanagan, K. J., Copland, D. A., Chenery, H. J., Byrne, G. J., & Angwin, A. J. (2013). Alzheimer’s disease is associated with distinctive semantic feature loss. Neuropsychologia, 51, 2016–2025. doi:10.1016/j.neuropsychologia.2013.06.008 Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). Mini-mental state: A practical guide for grading the mental state of patients for the clinician. Journal of Psychiatric Research, 12, 189–198. doi:10.1016/0022-3956(75)90026-6 Forbes-McKay, K. E., Ellis, A. W., Shanks, M. F., & Venneri, A. (2005). The age of acquisition of words produced in a semantic fluency task is highly predictive of early Alzheimer’s disease. Neuropsychologia, 43, 1128–1137. doi:10.1016/j.neuropsychologia.2006.07.040 Garrard, P., Lambon Ralph, M. A., Patterson, K., Pratt, K. H., & Hodges, J. R. (2005). Semantic feature knowledge and picture naming in dementia of Alzheimer’s type: A new approach. Brain and Language, 93, 79–94. doi:10.1016/j.bandl.2004.08.003 Glosser, G., Kohn, S. E., Friedman, R. B., Sands, L., & Grugan, P. K. (1997). Repetition of single words and nonwords in Alzheimer’s disease. Cortex, 33, 653–666. doi:10.1016/S0010-9452(08)70723-8 Gonzalez-Nosti, M., Barb on, A., Rodrıguez-Ferreiro, J., & Cuetos, F. (2013). Effects of psycholinguistic variables on the lexical decision task in Spanish: A study with 2,765 words. Behavior Research Methods, 46, 517–535. doi:10.3758/s13428-01300383-5 Hodges, J. R., Salmon, D. P., & Butters, N. (1991). The nature of the naming deficit in Alzheimer’s disease and Huntington’s disease. Brain, 114, 1547–1558. doi:10.1093/brain/113.3.1547 Hodges, J. R., Salmon, D. P., & Butters, N. (1992). Semantic memory impairment in Alzheimer’s disease: Failure of access or degraded knowledge? Neuropsychologia, 30, 301–314. doi:10.1093/brain/114.4.1547 Hoffman, P., Jones, R. W., & Lambon Ralph, M. A. (2013). Be concrete to be comprehended: Consistent imageability effects in semantic dementia for nouns, verbs, synonyms and associates. Cortex, 49, 1206–1218. doi:10.1016/j.cortex.2012.05.007
Word recognition in Alzheimer’s disease
13
Holmes, S. J., Fitch, F. J., & Ellis, A. W. (2006). Age of acquisition affects object recognition and picture naming in patients with Alzheimer’s disease. Journal of Clinical and Experimental Neuropsychology, 28, 1010–1022. doi:10.1080/13803390591004392 Huff, F. J., Corkin, S., & Growden, J. H. (1986). Semantic impairment and anomia in Alzheimer’s disease. Brain and Language, 28, 235–249. doi:10.1016/0093-934Z(86)90103-3 Irvine, K., Laws, K. R., Gale, T. M., & Kondel, T. K. (2012). Greater cognitive deterioration in women than men with Alzheimer’s disease: A meta analysis. Journal of Clinical and Experimental Neuropsychology, 34, 989–998. doi:10.1080/13803395.2012.712676 Kuperman, V., Schreuder, R., Bertram, R., & Baayen, R. H. (2009). Reading polymorphemic Dutch compounds: Toward a multiple route model of lexical processing. Journal of Experimental Psychology: Human Perception and Performance, 35, 876–895. doi:10.1037/a0013484 Lambon Ralph, M. A., Patterson, K. E., & Hodges, J. R. (1997). The relationship between naming and semantic knowledge for different categories in dementia of Alzheimer’s type. Neuropsychologia, 35, 1251–1260. doi:10.1016/S0028-3932(97)00052-3 Lambon Ralph, M. A., Powell, J., Howard, D., Whitworth, A. B., Garrard, P., & Hodges, J. R. (2001). Semantic memory is impaired in both dementia with Lewy bodies and dementia of Alzheimer’s type: A comparative neuropsychological study and literature review. Journal of Neurology, Neurosurgery and Psychiatry, 70, 149–156. doi:10.1136/jnnp.70.2.149 Law, R., & O’Carroll, R. E. (1998). A comparison of three measures for estimating premorbid intellectual level in dementia of the Alzheimer type. International Journal of Geriatric Psychiatry, 13, 727–730. doi:10.1002/(SICI)1099-1166(1998100)13 Madden, D. J., Welsh-Bohmer, K. A., & Tupler, L. A. (1999). Task complexity and signal detection analyses of lexical decision performance in Alzheimer’s disease. Developmental Neuropsychology, 16, 1–18. doi:10.1207/S15326942DN160101 Marczinski, C. A., & Kertesz, A. (2006). Category and letter fluency in semantic dementia, primary progressive aphasia, and Alzheimer’s disease. Brain and Language, 97, 258–265. doi:10.1016/ j.bandl.2005.11.001 Marquez, J. F., Cappa, S. F., & Sartori, G. (2011). Naming from definition, semantic relevance and feature type. Neuropsychology, 25, 105–113. doi:10.1037/a0020417 Martin, A., & Fedio, P. (1983). Word production and comprehension in Alzheimer’s disease: The breakdown of semantic knowledge. Brain and Language, 19, 124–141. doi:10.1016/0093934X(83)90059-7 McKhann, G., Drachman, D., Folstein, M., Katzman, R., Price, D., & Stadlan, E. M. (1984). Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA work group under the auspices of the Department of Health and Human Services Task Force on Alzheimer’s disease. Neurology, 34, 939–944. doi:10.1212/WNL.34.7.939 McKhann, G., Knopman, D. S., Chertkow, H., Hyman, B. T., Jack, Jr, C. R., Kawas, C. H., ... & Phelps, C. H. (2011). The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging and the Alzheimer’s Association workgroup. Alzheimer’s and Dementia, 7, 263–269. Perez, M. A., Alameida, J. R., & Cuetos, F. (2003). Frecuencia, longitud y vecindad ortografica de las palabras de 3 a 16 letras del diccionario de la lengua espa~ nola (rae, 1992). Revista Electr onica de Metodologıa Aplicada, 8, 1–20. Perri, R., Zannino, G., Caltagirone, C., & Carlesimo, G. A. (2012). Alzheimer’s disease and semantic deficits: A feature-listing study. Neuropsychology, 26, 652–663. doi:10.1037/a0029302 Peters, F., Majerus, S., De Baerdemaeker, J., Salmon, E., & Collette, F. (2009). Impaired semantic knowledge underlies the reduced verbal short-term storage capacity in Alzheimer’s disease. Neuropsychologia, 47, 3067–3073. doi:10.1016/j.neuropsychologia.2009.07.002 R Development Core Team (2012). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from http://www.Rproject.org/. ISBN 3-900051-07-0.
14
Fernando Cuetos et al.
Rissenberg, M., & Glanzer, M. (1987). Free recall and word finding ability in normal aging and senile dementia of the Alzheimer’s type: The effect of item concreteness. Journal of Gerontology, 42, 318–322. doi:10.1093/geronj/42.3.318 Rodrıguez-Ferreiro, J., Davies, R., Gonzalez-Nosti, M., Barb on, A., & Cuetos, F. (2009). Name agreement, frequency and age of acquisition, but not grammatical class, affect object and action naming in Spanish speaking participants with Alzheimer’s disease. Journal of Neurolinguistics, 22, 37–54. doi:10.1016/j.jneuroling.2008.05.003 Sailor, K. M., Antoine, M., Diaz, M., Kuslansky, G., & Kluger, A. (2004). The effects of Alzheimer’s disease on item output in fluency tasks. Neuropsychology, 18, 306–314. doi:10.1037/0894410518.2.306 Sailor, K. M., Zimmerman, M. E., & Sanders, A. E. (2011). Differential impacts of age of acquisition on letter and semantic fluency in Alzheimer’s disease patients and healthy older controls. Quarterly Journal of Experimental Psychology, 64, 2382–2391. doi:10.1080/17470218.2011.596660 Sebastian, N., Martı, M. A., Carreiras, M. F., & Cuetos, F. (2000). LEXESP: L exico informatizado del ~ Barcelona, Spain: Universidad de Barcelona. espanol. Silveri, M. C., Cappa, A., Mariotti, P., & Puopolo, M. (2002). Naming in patients with Alzheimer’s disease: Influence of age of acquisition and categorical effects. Journal of Clinical and Experimental Neuropsychology, 24, 755–764. doi:10.1076/jcen.24.6.755.8407 Taler, V., & Phillips, N. A. (2008). Language performance in Alzheimer’s disease and mild cognitive impairment: A comparative review. Journal of Clinical and Experimental Neuropsychology, 30, 501–556. doi:10.1080/13803390701550128 Tippett, L. J., Meier, S. L., Blackwood, K., & Diaz-Asper, C. (2007). Category-specific deficits in Alzheimer’s disease: Fact or artefact? Cortex, 43, 907–920. doi:10.1016/S0010-9452(08)70690-7 Verma, M., & Howard, R. J. (2012). Semantic memory and language dysfunction in early Alzheimer’s disease: A review. International Journal of Geriatric Psychiatry, 27, 1209–1217. doi:10.1002/ gps.3766 Received 13 July 2014; revised version received 13 May 2015
Journal of Neuropsychology (2015) © 2015 The British Psychological Society www.wileyonlinelibrary.com
Word recognition in Alzheimer’s disease: Effects of semantic degeneration Fernando Cuetos1, Noemı Arce1, Carmen Martınez2 and Andrew W. Ellis3,* 1
University of Oviedo, Spain Cabue~ nes Hospital, Gij on, Spain 3 University of York, UK 2
Impairments of word recognition in Alzheimer’s disease (AD) have been less widely investigated than impairments affecting word retrieval and production. In particular, we know little about what makes individual words easier or harder for patients with AD to recognize. We used a lexical selection task in which participants were shown sets of four items, each set consisting of one word and three non-words. The task was simply to point to the word on each trial. Forty patients with mild-to-moderate AD were significantly impaired on this task relative to matched controls who made very few errors. The number of patients with AD able to recognize each word correctly was predicted by the frequency, age of acquisition, and imageability of the words, but not by their length or number of orthographic neighbours. Patient Mini-Mental State Examination and phonological fluency scores also predicted the number of words recognized. We propose that progressive degradation of central semantic representations in AD differentially affects the ability to recognize low-imageability, low-frequency, lateacquired words, with the same factors affecting word recognition as affecting word retrieval.
Two of the cognitive problems most commonly reported in the early stages of Alzheimer’s disease (AD) are difficulty remembering recent events (episodic memory) and difficulty understanding and producing words (semantic memory; see Altmann & McClung, 2008; Lambon Ralph et al., 2001; Taler & Phillips, 2008; Verma & Howard, 2012). Problems with word retrieval have been studied intensively through the use of picture naming and other tasks (Chertkow & Bub, 1990; Chertkow, Bub, & Caplan, 1992; Cuetos, Martinez, Martinez, Izura, & Ellis, 2003; Hodges, Salmon, & Butters, 1991, 1992; Martin & Fedio, 1983). Problems with word recognition and comprehension in AD have also been documented and, like the problems in word retrieval, have mainly been attributed to impairments affecting the mental representations of concepts, including the semantic representations of familiar objects. Deficits in word recognition remain, however, underinvestigated compared with deficits in word retrieval and production. Evidence for a semantic basis to the lexical problems in AD includes the following: (1) the ability of patients to name specific objects is predicted by the amount of semantic
*Correspondence should be addressed to Andrew W. Ellis, Department of Psychology, University of York, York YO10 5DD, UK (email: [email protected]). DOI:10.1111/jnp.12077
2
Fernando Cuetos et al.
information they retain about those objects (Garrard, Lambon Ralph, Patterson, Pratt, & Hodges, 2005; Lambon Ralph, Patterson, & Hodges, 1997); (2) semantic naming errors (such as naming penguin as duck, or cow as milk) predominate in the early stages of the disease (Barbarotto, Capitani, Jori, Laiacona, & Molinari, 1998; Cuetos, Gonzalez-Nosti, & Martınez, 2005; Cuetos, Rodrıguez-Ferreiro, Sage, & Ellis, 2012; Hodges et al., 1991; Huff, Corkin, & Growden, 1986; Martin & Fedio, 1983; Rodrıguez-Ferreiro, Davies, GonzalezNosti, Barb on, & Cuetos, 2009); and (3) distinctive semantic information about concepts is lost sooner than information that is shared with many other concepts (Flanagan, Copland, Chenery, Byrne, & Angwin, 2013) with the result that patients can generate fewer atypical members of particular categories than typical members (Perri, Zannino, Caltagirone, & Carlesimo, 2012; Sailor, Antoine, Diaz, Kuslansky, & Kluger, 2004). One of the issues addressed in this paper is the extent to which the same properties of words predict success or failure in word retrieval and word recognition by patients with AD. Two factors that have emerged as consistent predictors of object naming accuracy in AD are word frequency and age of acquisition (AoA), with patients being better at retrieving and producing high-frequency names and names learned early in life than lowfrequency names and names learned later in life (Cuetos, Rodrıguez-Ferreiro et al., 2012; Rodrıguez-Ferreiro et al., 2009; Tippett, Meier, Blackwood, & Diaz-Asper, 2007). Theoretical interpretations of those effects have proposed that the semantic representations of concepts learned early in life and activated with high frequency thereafter are richer and/or easier to access and that the semantic representations of concepts are learned later in life and activated less often. Richer, more accessible semantic representations survive the early stages of AD better than poorer, less accessible representations and are, for example, able to drive successful name retrieval in patients in the mild-to-moderate stages of AD (see Ellis, 2011; for a review). The apparent lack of influence of the complexity of object pictures or the length of object names on picture naming in patients with AD has been interpreted as indicating that the primary impairment is semantic rather than, for example, perceptual or phonological (Albanese, 2007; Cuetos et al., 2005; Cuetos, Rodrıguez-Ferreiro et al., 2012; Rodrıguez-Ferreiro et al., 2009; Silveri, Cappa, Mariotti, & Puopolo, 2002; Tippett et al., 2007). One problem with investigating word recognition in AD and other neuropsychological conditions is that the tasks used to assess recognition and comprehension often involve presenting more than one word or picture at a time; for example, asking patients to point to the picture that matches a target word from a set of alternatives, or indicate whether pairs of words have similar or dissimilar meanings. When multiple words or pictures are presented, it becomes difficult to separate the properties of the target words from the properties of the other words or pictures that accompany them. The lexical decision task has sometimes been employed in an effort to overcome these problems. In lexical decision, participants are shown a sequence of words interleaved with invented nonwords (i.e., strings of letters that look like words but happen not to be; for example, flupe or quentole). In the traditional form of the lexical decision task, one word or non-word is presented at a time and the participant is required to indicate whether each item is a word or not. There is no requirement to demonstrate any understanding of the meanings of the words presented. Using visual presentation of written words and non-words, Chertkow and Bub (1989, 1990), Chertkow et al. (1992), Cuetos et al. (2003) and Du~ nabeitia, Marın, and Carreiras (2009) reported that patients with AD showed similar levels of accuracy in lexical decision as healthy, matched controls. In contrast, Madden, WelshBohmer, and Tupler (1999) reported less accurate lexical decision in patients with AD
Word recognition in Alzheimer’s disease
3
using written presentation while Glosser, Kohn, Friedman, Sands, and Grugan (1997) reported less accurate performance when the stimuli were spoken words and non-words. We will argue here that patients with AD are impaired at lexical decision relative to controls because they fail to recognize some words with which they were once familiar (and which matched controls continue to recognize). We believe that there are at least three reasons why previous studies have sometimes failed to observe reduced levels of accuracy in patients with AD. First, some studies have used object names as the words in their lexical decision tasks (Chertkow & Bub, 1989; Chertkow et al., 1992; Cuetos et al., 2003). By definition, object names are high in imageability. They tend also to be learned relatively early in life. If word recognition in patients with AD is better for highimageability, early-acquired words than for low-imageability, late-acquired words, then object names may be relatively easy for patients with AD to recognize and therefore not particularly well suited to identifying problems with word recognition. A second issue was highlighted by Madden et al. (1999) who found a decrease in lexical decision accuracy in patients with AD relative to controls that was greater for nonwords than for words. That is, patients appeared able to recognize most of the words correctly while miscategorising many non-words as words. This may indicate that when patients with AD are unsure as to whether an item is a word or a non-word, they tend to categorize it as a word. A Yes bias of this sort will sustain a spuriously high level of apparently correct responses to words and will only be revealed in the standard lexical decision task if overall accuracy among the patients with AD is not at ceiling and if error rates to non-words are also analysed. Madden et al. (1999) addressed this issue using a signal detection analysis which demonstrated a substantial bias towards responding Yes to non-words, but also an impaired sensitivity to words over and above the change in response bias. A simple way to eliminate response bias is to modify the lexical decision task to a version in which participants are presented on each trial with (for example) one word and one non-word and are asked to indicate which item is the real word. With a requirement to pick out the word on every trial, response biases are removed. Baddeley, Emslie, and Nimmo-Smith’s (1993) ‘Spot the Word’ test involved just such a two-alternative forcedchoice version of lexical decision (Baddeley & Crawford, 2012). Using that test, Law and O’Carroll (1998) found a trend towards more errors in patients with AD compared with controls, while Beardsall and Huppert (1997) reported more errors in patients with mildto-moderate AD than in patients with minimal dementia. But although these studies dealt with the issue of response bias, they illustrate a third potential problem. If lexical selection or lexical decision tasks use half words and half non-words, participants have a 50% chance of making a correct response on each trial even if they actually have no idea which of the stimuli are words and which are non-words. A 50% chance rate makes accuracy scores relatively insensitive to impairments. The lexical selection task can be made more sensitive by increasing the number of non-words that accompany each word target, thereby reducing the probability of selecting the real word purely by chance. Cuetos, Herrera, and Ellis (2010) presented three non-words with each real word in a lexical selection task, thereby reducing the chance of a fortuitously correct response to 25%. Patients with mild-to-moderate AD correctly selected the real word on an average of 88% of trials, while matched controls selected the word on an average of 99% of trials. Although well above chance, patients with AD performed significantly worse than controls. If the problems affecting word recognition in AD share a common origin with the problems affecting word retrieval (e.g., impairment to semantic representations affecting both word recognition and production), we would expect the same factors to influence
4
Fernando Cuetos et al.
lexical selection as have been shown to influence word retrieval. As noted above, one of the factors consistently found to influence word retrieval by patients with AD in naming tasks is AoA: All other things being equal, patients are better able to name objects with early- than late-acquired names (Albanese, 2007; Catling, South, & Dent, 2013; Cuetos, Rodrıguez-Ferreiro et al., 2012; Holmes, Fitch, & Ellis, 2006; Marquez, Cappa, & Sartori, 2011; Rodrıguez-Ferreiro et al., 2009; Silveri et al., 2002; Tippett et al., 2007). Effects of AoA have also been observed when patients are asked to generate words belonging to different semantic categories (Forbes-McKay, Ellis, Shanks, & Venneri, 2005; Sailor, Zimmerman, & Sanders, 2011). In addition, Perri et al. (2012) found that patients with mild AD were able to generate more semantic features for early- than late-acquired concepts. Cuetos, Rodrıguez-Ferreiro et al. (2012) found worse lexical selection performance for late- than early-acquired words in the lexical selection task using sets of words matched on frequency, imageability, and other factors. AoA would therefore appear to be one property of words and concepts that affects the accuracy of both word retrieval and word recognition in patients with AD. A number of recent studies have reported effects of word frequency on object naming in patients with AD that were independent of the effects of AoA and other factors (Cuetos, Rodrıguez-Ferreiro et al., 2012; Rodrıguez-Ferreiro et al., 2009; Tippett et al., 2007). We therefore predicted an effect of word frequency on lexical selection in patients with AD independent of the effects of any other factors. It is hard to evaluate the effects of imageability on object naming because objects tend to score highly on imageability, so the range of values available for manipulation is limited. Failures to find the effects of imageability on naming accuracy in patients with AD (Albanese, 2007; Cuetos et al., 2005; Cuetos, Rodrıguez-Ferreiro et al., 2012; Rodrıguez-Ferreiro et al., 2009) cannot form the basis of firm predictions regarding possible effects in lexical selection where a much wider range of imageability values can be employed. Two studies have, however, reported effects of imageability/concreteness in patients with AD using alternative tasks. Rissenberg and Glanzer (1987) reported that the impairment shown by patients with AD at producing words in response to definitions was substantially greater when the target words were abstract than when they were concrete while Peters, Majerus, De Baerdemaeker, Salmon, and Collette (2009) found better immediate serial recall of highthan low-imageability words in patients with AD. Effects of imageability (or concreteness) are generally regarded as indicators of semantic involvement in particular tasks, the assumption being made that the limited perceptual experience associated with abstract words causes their semantic representations to be less detailed than the representations of concrete words (Hoffman, Jones, & Lambon Ralph, 2013). If the underlying deficit in patients with AD is primarily semantic, and if imageability is a semantic variable, then we would predict an effect of imageability on lexical selection in the present study. Some words in the language resemble many other words, while others are more distinctive in their appearance. One measure of distinctiveness is orthographic neighbourhood size (N), defined as the number of other words that differ from a target word by a single letter (Coltheart, Davelaar, Jonasson, & Besner, 1977). Distinctive words have low N values, while typical words have higher values. The only study we are aware of that has investigated the effects of N on object naming in patients with AD is by RodrıguezFerreiro et al. (2009) who found no significant effect. Du~ nabeitia et al. (2009) compared lexical decision accuracy and reaction times to words with many or few orthographic neighbours in patients with AD and in controls. Both groups made more errors to low than high N words, but the difference in favour of the high N words was smaller in the patients with AD and not significant. That is, in terms of its effect on accuracy, the effect of N was
Word recognition in Alzheimer’s disease
5
reduced in patients with AD compared with controls. Du~ nabeitia et al. (2009) also reported comparable effects of N on reaction times in patients with AD and in controls. Accuracy of responses to non-words was not reported, so the possibility of a Yes bias in the patients cannot be excluded. Finally, studies of word retrieval in AD have consistently failed to find effects of word length. When word frequency and other factors are taken into account, patients are as likely to be able to retrieve and produce a long object name correctly as a short name (Cuetos, Rodrıguez-Ferreiro et al., 2012; Rodrıguez-Ferreiro et al., 2009; Silveri et al., 2002; Tippett et al., 2007). We are not aware of any previous investigations of the influence of length on word recognition in patients with AD, but if the principal deficit in AD is semantic rather than perceptual, orthographic, or phonological, there is no reason to expect an effect of length on lexical selection performance in patients with AD. The stimuli of interest in the present study were 150 words that varied on word frequency, AoA, imageability, N, and length. These were chosen to be words that the controls were expected to know. On each trial of the lexical selection task, one written word was accompanied by three non-words that were matched to the word on number of letters and syllables. Participants were simply asked to indicate which item in the array was a real word. The participants were 40 patients with mild-to-moderate AD and 25 matched controls. We predicted that the patients with AD would have difficulty recognizing the more low-frequency, late-acquired, and low-imageability words. We did not expect to find an effect of length. The literature was ambiguous as to whether an effect of N might be observed. In presenting the results, we report one set of correlations which show the relationships across patients between age, years of education, Mini-Mental State Examination (MMSE) scores, scores on two fluency tasks, and lexical selection accuracy. We then present an item analysis showing the correlations across words between frequency, AoA, imageability, etc. and the accuracy with which the patients with AD as a group responded to those words. The main statistical analysis is, however, a mixed-effects multiple regression. Where a conventional multiple regression would take a mean level of accuracy for each word across patients and use it to look at the effects of word properties like frequency and length, a mixed-effects multiple regression takes every response from every participant to every word (i.e., 40 patients with AD 9 150 words = 6,000 data points) and allows the researcher to analyse the effects of participant and stimulus characteristics in the same analysis which is more sensitive to effects than a traditional multiple regression (cf. Cuetos, Rodrıguez-Ferreiro et al., 2012; Gonzalez-Nosti, Barb on, Rodrıguez-Ferreiro, & Cuetos, 2013). Our mixed-effects analysis combined participant MMSE and fluency scores with frequency, AoA, and the other lexical variables.
Methods Participants Forty patients with probable dementia of Alzheimer’s type (35 female and 5 male aged 66– 91 years) took part in the study. The patients were selected on the basis of their medical history, information from a knowledgeable informant, a CT or MRI scan, and a neuropsychological evaluation which included the MMSE (Folstein, Folstein, & McHugh, 1975) along with tests of semantic fluency (naming as many animals as possible in 1 min) and phonological fluency (producing as many words beginning with ‘s’ as possible in 1 min). The diagnosis of probable AD was made according to the Neurological and
6
Fernando Cuetos et al.
Communicative Disorders and Stroke – Alzheimer’s Disease and Related Disorders (NINCDS-ARDA criteria; McKhann et al., 1984; revised by McKhann et al., 2011). Patients also met the DSM-IV criteria for dementia of the Alzheimer’s type (American Psychiatric Association, 2000). The study was approved by the Ethics Committee of the Hospital Cabue~ nes, Gij on. Informed consent was obtained from all participants and from patients’ caregivers prior to the study. The controls were 25 healthy adult volunteers (21 female and 4 male) matched to the patients on age and years of education (Table 1). None had a psychiatric history, sensory deficiencies, or medical conditions that could impair performance on the neuropsychological tests. Patients with AD differed significantly from the controls on MMSE score, t (63) = 10.62, p < .001, semantic fluency, t(63) = 8.77, p < .001, and phonological fluency, t(63) = 6.45, p < .001, but not on age, t(63) = 1.10, p = .276, or years of education, t(63) = 0.07, p = .946.
Materials One hundred and fifty Spanish words were selected for this study covering a range of values for AoA, word frequency, imageability, length, and N. The words were taken from the ratings study by Cuetos, Samartino, and Ellis (2012) and were items that a high proportion of adults in that study (30 healthy adults aged 61–85 years) recognized as familiar. AoA ratings were taken from the Cuetos, Samartino et al.’s (2012) study where they were obtained using a scale from 1 = learned before the age of 2 years to 8 = learned after the age of 20. Such ratings have been shown to correlate highly with objective measures of AoA (Brysbaert & Ellis, in press; Ellis, 2011). Word frequency values were taken from the Subtlex-Esp database (Cuetos, Gonzalez-Nosti, Barb on, & Brysbaert, 2011) which is based on the frequencies with which words occur in a corpus of 41.5 million words taken from contemporary film subtitles. Subtitle frequencies have been found to predict performance in word recognition experiments better than frequency counts based solely on written texts (Brysbaert & New, 2009). The words had a mean of 14.45 occurrences per million words of Spanish (range 0.19–311.85). Imageability values were taken from the LEXESP database (Sebastian, Martı, Carreiras, & Cuetos, 2000) where adult participants rated words on a 7-point scale from 1 = very hard to conjure up a mental image to 7 = very easy to conjure up a mental image. The words had a mean imageability value of 5.00 (range 2.07–6.67). Values of N were taken from Perez, Alameida, and Cuetos (2003). The words had a mean of 2.11 neighbours (range 0–17). Finally, word length was measured as the number of letters in each word Table 1. Summary of participant characteristics and scores on the MMSE, semantic fluency, and phonological fluency tasks Age Patients with AD Mean 78.70 SD 5.98 Controls Mean 77.20 SD 4.13
Years of education
MMSE
Semantic fluency
Phonological fluency
8.10 3.20
19.88 4.23
7.15 3.44
5.23 3.57
8.16 3.83
29.08 1.15
15.88 4.56
10.96 3.32
Note. AD = Alzheimer’s disease, MMSE = Mini-Mental State Examination, SD = standard deviation.
Word recognition in Alzheimer’s disease
7
(mean 6.15; range 4–9). During item selection, an effort was made to reduce the intercorrelations between the variables while maintaining a reasonable spread of values on each variable. Four hundred and fifty non-words were generated, three for each target word. The non-words were pronounceable, word-like letter strings which could be words in Spanish but happen not to be (e.g., lefa, milusa, and pasorento). The three non-words that accompanied each target word were matched to the target on the number of letters and syllables. To check that there were no confounds between the properties of the target words and the characteristics of the non-words that accompanied them, we calculated the mean length, N, and bigram frequencies of the non-words used on each trial. The correlations between the mean length, N, and bigram frequencies of the non-words and the frequency, AoA, and imageability of the accompanying words were all non-significant. It is unlikely, therefore, that any effect of frequency, AoA, or imageability found for the recognition of real words was an indirect consequence of differences in the ease or difficulty of rejecting the non-words that were presented with the words on each trial. The three non-words and one word were presented on the screen of a laptop in a square formation (two above and two below) using black, lower case font on a white background. Words were distributed evenly across the four possible positions, with a real word occurring at each position either 37 or 38 times.
Procedure Participants were tested individually in a quiet room in the Neurology Unit of the Cabue~ nes Hospital, Gij on. The experimenter first explained the lexical selection task using several examples. When the experimenter was satisfied that the participant understood the task, the experimental stimuli were presented one set at a time. On each trial, participants were asked to look at the four alternatives and indicate which was the real word by pointing to it. Occasionally, participants recognized the real word quickly and named it rather than pointing to it. Those responses were accepted, with the patient being reminded of the requirement to point to the real words rather than name them. To minimize non-verbal cues, the experimenter stood to the side and slightly behind the patient, noting the response on each trial by marking the chosen item on a response sheet which showed the four alternatives. No feedback was given during the experiment. The MMSE and fluency tasks were administered on the same day as the lexical selection task.
Results As expected, performance of the controls on the lexical selection task was close to ceiling. Controls identified a mean of 147.4 of the 150 words correctly (98.3%), with 20 of the 25 controls responding correctly to 147 or more of the words. The patients with AD identified a mean of 123.2 words correctly (82.1%). That is well above the chance rate of 25% but significantly worse than the accuracy of the controls, t(63) = 6.46, p < .001. Having established that the patients with AD were impaired on the lexical selection task, further analyses focused on the patient data. Table 2 shows the correlations between age, years of education, MMSE, semantic fluency, phonological fluency, and lexical selection for the patients with AD. Lexical selection scores correlated significantly with number of years of education, MMSE, semantic fluency, and phonological fluency.
8
Fernando Cuetos et al.
Table 2. Correlations between age, years of education, MMSE, semantic fluency, phonological fluency, and lexical selection for the patients with AD
Age Years of education MMSE Semantic fluency Phonological fluency Lexical selection
Age
Years of Educ.
MMSE
Sem. fluency
–
.307 –
.064 .156 –
.015 .082 .589*** –
Phon. fluency .048 .107 .410** .605*** –
Lexical selection .036 .377* .688*** .433** .455** –
Notes. Educ. = education, MMSE = Mini-Mental State Examination, Sem. = Semantic, Phon. = Phonological. *p < .05; **p < .01; ***p < .001.
Table 3 shows the correlations across items between the word properties (frequency, AoA, etc.) and the number of patients with AD who selected each word correctly. Lexical selection scores for the 150 words correlated significantly with word frequency, AoA, and imageability, but not with N or length. When the predictor variables are correlated with each other as well as with the dependent variable (number of patients selecting each word correctly), some form of regression analysis is required to determine which predictor variables are exerting genuinely independent effects on the dependent variable. We employed a mixed-effects multiple regression model implemented through the lme4 package in R (R Development Core Team, 2012, version 2.15). The model had patients and items as random intercepts and the predictors as fixed effect factors or covariates (cf. Baayen, Davidson, & Bates, 2008; Kuperman, Schreuder, Bertram, & Baayen, 2009). The patient characteristics employed as predictors were MMSE scores and scores on the semantic and phonological fluency tasks. The lexical predictors were word frequency, AoA, imageability, length, and N. Log values of the predictors were used to reduce skew. Collinearity was assessed by calculating the variance inflation factor (VIF) for each predictor. VIF provides a measure of how much larger the variance of a particular coefficient is than it would have been if that predictor was completely uncorrelated with the other predictors, with VIF values >4
Table 3. Correlations among the predictor variables and between the predictor variables and the number of patients with AD who selected each word correctly in the lexical selection test AoA AoA Frequency Imageability Length N Lexical selection
–
Freq
Imag
.281*** –
.531*** .029 –
Length .143 .029 .152 –
N .244** .099 .121 .648*** –
Lexical selection .431*** .234** .344*** .057 .134 –
Notes. AoA = age of acquisition, Freq = Frequency, Imag = Imageability, N = Number of orthographic neighbours. *p < .05; **p < .01; ***p < .001.
Word recognition in Alzheimer’s disease
9
generally taken to indicate potential problems with multicollinearity (Cohen, Cohen, West, & Aiken, 2003). The VIF values for the predictors in this study ranged from 1.275 to 1.714, indicating that multicollinearity was not a problem and that the analysis was valid. The analysis found significant effects of MMSE, Z = 4.55, p < .001, and phonological fluency, Z = 2.25, p < .05, among the participant variables and word frequency, Z = 3.28, p < .001, AoA, Z = 6.21, p < 001, and imageability, Z = 4.26, p < .001, among the lexical variables. In the context of these other predictors, the effects of semantic fluency score, Z = 0.52, p = .600, neighbourhood size (N), Z = 0.39, p = .692, and word length, Z = 0.56, p = .574, did not approach significance.
Discussion The words employed in the present study were chosen to be ones that the healthy controls were expected to recognize (and therefore, by inference, words that the patients with AD would have been able to recognize before the onset of their illness). As expected, control performance was at or close to ceiling while the patients with AD identified significantly fewer real words than the controls. We note that the patients with AD in our study were predominantly female. That would be expected from the longer life expectancy of women, but a meta-analysis by Irvine, Laws, Gale, and Kondel (2012) found that cognitive functions are more severely and more widely affected in women with AD than in men, even allowing for differences in age, education, or dementia severity. The controls in the present study had the same proportion of men and women as the patients with AD. The lexical selection task eliminates the problem of response bias that can affect traditional lexical decision and reduces the probability of making a correct response by chance. The finding that lexical selection was impaired in the patients with AD supports previous demonstrations of impaired lexical decision/selection in AD (Beardsall & Huppert, 1997; Cuetos et al., 2010; Glosser et al., 1997; Madden et al., 1999) and suggests that reports of near-normal lexical decision accuracy in patients with AD may have been due to a combination of response biases, high chance rates, and the use of relatively easy, high-imageability words. Having established that the patients with AD failed to recognize words they would previously have known, further analyses focused on the patient data. The mixed-effects regression analysis found that lexical selection accuracy was predicted by the MMSE and phonological fluency scores of the patients and by the frequency, AoA, and imageability of the words. In the context of those predictors, the effects of semantic fluency scores and the length and N values of words were not significant. Worse performance at word recognition in more severe patients with lower MMSE scores is to be expected (Cuetos et al., 2010). The effect of AoA, with better recognition of early- than late-acquired words, replicates the finding of Cuetos et al. (2010). The demonstration of an additional and independent effect of word frequency is new in this context, but independent effects of frequency and AoA have been observed in studies of object naming by patients with AD (Cuetos, Samartino et al., 2012; Rodrıguez-Ferreiro et al., 2009; Tippett et al., 2007). The observed effect of imageability is in line with reported effects of that variable in other tasks (Peters et al., 2009; Rissenberg & Glanzer, 1987). Taken together, the effects of frequency, AoA, and imageability are consistent with the hypothesis that a central semantic impairment underlies much of the difficulty that patients have in both word recognition and production.
10
Fernando Cuetos et al.
We chose words for our lexical selection task that we expected healthy controls to know because we see little point in demonstrating that patients with AD are unable to recognize little-known words that controls are also unable to recognize. A connection might nevertheless be made between impaired word recognition in patients with AD and normal word recognition in healthy adults. The connection is that the same factors that cause patients with AD to be unable to recognize once-familiar words also appear to affect the speed with which healthy adults can recognize words. Effects of word frequency, AoA, and imageability have been reported on the speed with which healthy adults respond correctly to words in lexical decision (Cortese & Khanna, 2007; Cortese & Schock, 2013; Gonzalez-Nosti et al., 2013). This suggests a general rule whereby words that healthy adults recognize correctly but slowly are the words that patients with AD fail to recognize at all. We propose that the mechanism underlying these effects is that impoverished semantic representations in healthy adults mean that low-frequency, lateacquired, and low-imageability words are recognized relatively slowly in lexical decision. Those impoverished representations are more vulnerable to the degenerative effects of AD than are the richer representations of high-frequency, early-acquired, and highimageability words. Words whose semantic representations have degenerated beyond a certain point no longer seem familiar and cannot be easily distinguished from non-words in the lexical selection task (Ellis, 2011). We found no effect of orthographic neighbourhood size on lexical selection accuracy in our AD group. Studies of object naming in patients with AD have also failed to find the effects of N (Cuetos, Samartino et al., 2012; Rodrıguez-Ferreiro et al., 2009; Tippett et al., 2007). In contrast, Du~ nabeitia et al. (2009) found the effects of N on lexical decision speed and accuracy both in patients with AD and in controls. Findings from other studies have been mixed with respect to the effects of N on lexical decision speed in healthy adults (Cortese & Khanna, 2007; Cortese & Schock, 2013; GonzalezNosti et al., 2013). Balota, Cortese, Sergent-Marshall, Spieler, and Yapp (2004) found no effects of N on lexical decision speed or accuracy in young adults but effects on both speed and accuracy in older adults. This is an issue where further research is needed, including exploring different measures of the similarity between word forms. Despite this apparent anomaly, the general rule that slow recognition in healthy adults converts into recognition failure in patients with AD is, we believe, supported by the available evidence. In addition to finding no significant effects of word length and N on lexical selection accuracy in our patient group, we found no effect of semantic fluency scores. Studies have indicated that the probability that patients with AD will retrieve particular words in the semantic fluency task is influenced by both the frequency and AoA of those words (Binetti et al., 1995; Forbes-McKay et al., 2005; Marczinski & Kertesz, 2006; Sailor et al., 2011). It may be that when frequency and AoA are included as predictors of word recognition accuracy along with MMSE and phonological fluency, little or no variance remains for semantic fluency to account for.
Acknowledgements This research was supported by grant PSI2012-31913 from the Spanish Government.
Word recognition in Alzheimer’s disease
11
References Albanese, E. (2007). The “hidden” semantic category dissociation in mild–moderate Alzheimer’s disease patients. Neuropsychologia, 45, 639–643. doi:10.1016/j.neuropsychologia.2006.07.018 Altmann, L. J. P., & McClung, J. S. (2008). Effects of semantic impairment on language use in Alzheimer’s disease. Seminars in Speech and Language, 29, 18–31. doi:10.1055/s-20081061622 American Psychiatric Association (2000). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author. Baayen, R. H., Davidson, D. J., & Bates, D. M. (2008). Mixed-effects modeling with crossed random effects for subjects and items. Journal of Memory and Language, 59, 390–412. doi:10.1016/ jml.2007.12.005 Baddeley, A. D., & Crawford, J. (2012). Spot the word (2nd ed.). Oxford, UK: Pearson. Baddeley, A. D., Emslie, H., & Nimmo-Smith, I. N. (1993). The Spot-the-Word test: A robust estimate of verbal intelligence based on lexical decision. British Journal of Clinical Psychology, 32, 55– 66. doi:10.1111/j.2044-8260.1993.tb01027.x Balota, D. A., Cortese, M. J., Sergent-Marshall, S. D., Spieler, D. H., & Yapp, M. J. (2004). Visual word recognition for single-syllable words. Journal of Experimental Psychology: General, 133, 283– 316. doi:10.1037/0096-3445.133.2.283 Barbarotto, R., Capitani, E., Jori, T., Laiacona, M., & Molinari, S. (1998). Picture naming and progression of Alzheimer’s disease: An analysis of error types. Neuropsychologia, 36, 397–405. doi:10.1016/S0028-3932(97)00124-3 Beardsall, L., & Huppert, F. A. (1997). Short NART, CCRT, and Spot-the-Word: Comparisons in older and demented persons. British Journal of Clinical Psychology, 36, 619–622. doi:10.1111/ j.2044-8260.1997.tb01266.x Binetti, G., Magni, E., Cappa, S. F., Padovani, A., Bianchetti, A., & Trabucchi, M. (1995). Semantic memory in Alzheimer’s disease: An analysis of category fluency. Journal of Clinical and Experimental Neuropsychology, 17, 82–89. doi:10.1080/13803399508406584 Brysbaert, M., & Ellis, A. W. (in press). Aphasia and age of acquisition: Are early-learned words more resilient? Aphasiology. Brysbaert, M., & New, B. (2009). Moving beyond Kucera and Francis: A critical evaluation of current word frequency norms and the introduction of a new and improved word frequency measure for American English. Behavior Research Methods, 41, 977–990. doi:10.3758/BRM.41.4.977 Catling, J., South, F., & Dent, K. (2013). The effect of age of acquisition on older individuals with and without cognitive impairments. Quarterly Journal of Experimental Psychology, 66, 1963– 1973. doi:10.1080/17470218.2013.771689 Chertkow, H., & Bub, D. (1989). Priming and semantic memory loss Alzheimer’s disease. Brain and Language, 36, 420–446. doi:10.1016/0093-934X(89)90078-3 Chertkow, H., & Bub, D. (1990). Semantic memory loss in dementia of Alzheimer’s type. Brain, 113, 397–417. doi:10.1093/brain/113.2.397 Chertkow, H., Bub, D., & Caplan, D. (1992). Constraining theories of semantic memory processing: Evidence from dementia. Cognitive Neuropsychology, 9, 327–365. doi:10.1080/ 02643299208252064 Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: Lawrence Erlbaum. Coltheart, M., Davelaar, E., Jonasson, J. T., & Besner, D. (1977). Access to the internal lexicon. In S. Dornic (Ed.), Attention and performance, VI (pp. 535–555). New York, NY: Academic Press. Cortese, M. J., & Khanna, M. M. (2007). Age of acquisition predicts naming and lexical-decision performance above and beyond 22 other predictor variables: An analysis of 2,342 words. Quarterly Journal of Experimental Psychology, 60, 1072–1082. doi:10.1080/1747021070 1315467
12
Fernando Cuetos et al.
Cortese, M. J., & Schock, J. (2013). Imageability and age of acquisition effects in disyllabic word recognition. Quarterly Journal of Experimental Psychology, 66, 946–972. doi:10.1080/ 17470218.2012.722660 Cuetos, F., Gonzalez-Nosti, M., Barb on, A., & Brysbaert, M. (2011). SUBTLEX-ESP: Spanish word frequencies based on film subtitles. Psicol ogica, 32, 133–143. Cuetos, F., Gonzalez-Nosti, M., & Martınez, C. (2005). The picture-naming task in the analysis of cognitive deterioration in Alzheimer’s disease. Aphasiology, 19, 545–557. doi:10.1080/ 02687030544000010 Cuetos, F., Herrera, E., & Ellis, A. W. (2010). Impaired word recognition in Alzheimer’s disease: The role of age of acquisition. Neuropsychologia, 48, 3329–3334. doi:10.1016/j.neuropsychologia.2012.07.017 Cuetos, F., Martinez, T., Martinez, C., Izura, C., & Ellis, A. W. (2003). Lexical processing in Spanish patients with probable Alzheimer’s disease. Cognitive Brain Research, 17, 549–561. doi:10.1016/S0926=6410(03)00169-1 Cuetos, F., Rodrıguez-Ferreiro, J., Sage, K., & Ellis, A. W. (2012). A fresh look at the predictors of naming accuracy and errors in Alzheimer’s disease. Journal of Neuropsychology, 6, 242–256. doi:10.1111/j.1748-6653.2011.02025.x Cuetos, F., Samartino, T., & Ellis, A. W. (2012). Age acquisition norms from elderly Spanish people: Characteristics and the prediction of word recognition performance in Alzheimer’s disease. Psicol ogica, 33, 59–76. Du~ nabeitia, J. A., Marın, A., & Carreiras, M. (2009). Associative and orthographic neighborhood density effects in normal aging and Alzheimer’s disease. Neuropsychology, 23, 759–764. doi:10.1037/a0016616 Ellis, A. W. (2011). The acquisition, retention and loss of vocabulary in aphasia, dementia and other neuropsychological conditions. In M. Faust (Ed.), Handbook of the neuropsychology of language (Vol. 2, pp. 637–660). Oxford, UK: Blackwells. Flanagan, K. J., Copland, D. A., Chenery, H. J., Byrne, G. J., & Angwin, A. J. (2013). Alzheimer’s disease is associated with distinctive semantic feature loss. Neuropsychologia, 51, 2016–2025. doi:10.1016/j.neuropsychologia.2013.06.008 Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). Mini-mental state: A practical guide for grading the mental state of patients for the clinician. Journal of Psychiatric Research, 12, 189–198. doi:10.1016/0022-3956(75)90026-6 Forbes-McKay, K. E., Ellis, A. W., Shanks, M. F., & Venneri, A. (2005). The age of acquisition of words produced in a semantic fluency task is highly predictive of early Alzheimer’s disease. Neuropsychologia, 43, 1128–1137. doi:10.1016/j.neuropsychologia.2006.07.040 Garrard, P., Lambon Ralph, M. A., Patterson, K., Pratt, K. H., & Hodges, J. R. (2005). Semantic feature knowledge and picture naming in dementia of Alzheimer’s type: A new approach. Brain and Language, 93, 79–94. doi:10.1016/j.bandl.2004.08.003 Glosser, G., Kohn, S. E., Friedman, R. B., Sands, L., & Grugan, P. K. (1997). Repetition of single words and nonwords in Alzheimer’s disease. Cortex, 33, 653–666. doi:10.1016/S0010-9452(08)70723-8 Gonzalez-Nosti, M., Barb on, A., Rodrıguez-Ferreiro, J., & Cuetos, F. (2013). Effects of psycholinguistic variables on the lexical decision task in Spanish: A study with 2,765 words. Behavior Research Methods, 46, 517–535. doi:10.3758/s13428-01300383-5 Hodges, J. R., Salmon, D. P., & Butters, N. (1991). The nature of the naming deficit in Alzheimer’s disease and Huntington’s disease. Brain, 114, 1547–1558. doi:10.1093/brain/113.3.1547 Hodges, J. R., Salmon, D. P., & Butters, N. (1992). Semantic memory impairment in Alzheimer’s disease: Failure of access or degraded knowledge? Neuropsychologia, 30, 301–314. doi:10.1093/brain/114.4.1547 Hoffman, P., Jones, R. W., & Lambon Ralph, M. A. (2013). Be concrete to be comprehended: Consistent imageability effects in semantic dementia for nouns, verbs, synonyms and associates. Cortex, 49, 1206–1218. doi:10.1016/j.cortex.2012.05.007
Word recognition in Alzheimer’s disease
13
Holmes, S. J., Fitch, F. J., & Ellis, A. W. (2006). Age of acquisition affects object recognition and picture naming in patients with Alzheimer’s disease. Journal of Clinical and Experimental Neuropsychology, 28, 1010–1022. doi:10.1080/13803390591004392 Huff, F. J., Corkin, S., & Growden, J. H. (1986). Semantic impairment and anomia in Alzheimer’s disease. Brain and Language, 28, 235–249. doi:10.1016/0093-934Z(86)90103-3 Irvine, K., Laws, K. R., Gale, T. M., & Kondel, T. K. (2012). Greater cognitive deterioration in women than men with Alzheimer’s disease: A meta analysis. Journal of Clinical and Experimental Neuropsychology, 34, 989–998. doi:10.1080/13803395.2012.712676 Kuperman, V., Schreuder, R., Bertram, R., & Baayen, R. H. (2009). Reading polymorphemic Dutch compounds: Toward a multiple route model of lexical processing. Journal of Experimental Psychology: Human Perception and Performance, 35, 876–895. doi:10.1037/a0013484 Lambon Ralph, M. A., Patterson, K. E., & Hodges, J. R. (1997). The relationship between naming and semantic knowledge for different categories in dementia of Alzheimer’s type. Neuropsychologia, 35, 1251–1260. doi:10.1016/S0028-3932(97)00052-3 Lambon Ralph, M. A., Powell, J., Howard, D., Whitworth, A. B., Garrard, P., & Hodges, J. R. (2001). Semantic memory is impaired in both dementia with Lewy bodies and dementia of Alzheimer’s type: A comparative neuropsychological study and literature review. Journal of Neurology, Neurosurgery and Psychiatry, 70, 149–156. doi:10.1136/jnnp.70.2.149 Law, R., & O’Carroll, R. E. (1998). A comparison of three measures for estimating premorbid intellectual level in dementia of the Alzheimer type. International Journal of Geriatric Psychiatry, 13, 727–730. doi:10.1002/(SICI)1099-1166(1998100)13 Madden, D. J., Welsh-Bohmer, K. A., & Tupler, L. A. (1999). Task complexity and signal detection analyses of lexical decision performance in Alzheimer’s disease. Developmental Neuropsychology, 16, 1–18. doi:10.1207/S15326942DN160101 Marczinski, C. A., & Kertesz, A. (2006). Category and letter fluency in semantic dementia, primary progressive aphasia, and Alzheimer’s disease. Brain and Language, 97, 258–265. doi:10.1016/ j.bandl.2005.11.001 Marquez, J. F., Cappa, S. F., & Sartori, G. (2011). Naming from definition, semantic relevance and feature type. Neuropsychology, 25, 105–113. doi:10.1037/a0020417 Martin, A., & Fedio, P. (1983). Word production and comprehension in Alzheimer’s disease: The breakdown of semantic knowledge. Brain and Language, 19, 124–141. doi:10.1016/0093934X(83)90059-7 McKhann, G., Drachman, D., Folstein, M., Katzman, R., Price, D., & Stadlan, E. M. (1984). Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA work group under the auspices of the Department of Health and Human Services Task Force on Alzheimer’s disease. Neurology, 34, 939–944. doi:10.1212/WNL.34.7.939 McKhann, G., Knopman, D. S., Chertkow, H., Hyman, B. T., Jack, Jr, C. R., Kawas, C. H., ... & Phelps, C. H. (2011). The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging and the Alzheimer’s Association workgroup. Alzheimer’s and Dementia, 7, 263–269. Perez, M. A., Alameida, J. R., & Cuetos, F. (2003). Frecuencia, longitud y vecindad ortografica de las palabras de 3 a 16 letras del diccionario de la lengua espa~ nola (rae, 1992). Revista Electr onica de Metodologıa Aplicada, 8, 1–20. Perri, R., Zannino, G., Caltagirone, C., & Carlesimo, G. A. (2012). Alzheimer’s disease and semantic deficits: A feature-listing study. Neuropsychology, 26, 652–663. doi:10.1037/a0029302 Peters, F., Majerus, S., De Baerdemaeker, J., Salmon, E., & Collette, F. (2009). Impaired semantic knowledge underlies the reduced verbal short-term storage capacity in Alzheimer’s disease. Neuropsychologia, 47, 3067–3073. doi:10.1016/j.neuropsychologia.2009.07.002 R Development Core Team (2012). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from http://www.Rproject.org/. ISBN 3-900051-07-0.
14
Fernando Cuetos et al.
Rissenberg, M., & Glanzer, M. (1987). Free recall and word finding ability in normal aging and senile dementia of the Alzheimer’s type: The effect of item concreteness. Journal of Gerontology, 42, 318–322. doi:10.1093/geronj/42.3.318 Rodrıguez-Ferreiro, J., Davies, R., Gonzalez-Nosti, M., Barb on, A., & Cuetos, F. (2009). Name agreement, frequency and age of acquisition, but not grammatical class, affect object and action naming in Spanish speaking participants with Alzheimer’s disease. Journal of Neurolinguistics, 22, 37–54. doi:10.1016/j.jneuroling.2008.05.003 Sailor, K. M., Antoine, M., Diaz, M., Kuslansky, G., & Kluger, A. (2004). The effects of Alzheimer’s disease on item output in fluency tasks. Neuropsychology, 18, 306–314. doi:10.1037/0894410518.2.306 Sailor, K. M., Zimmerman, M. E., & Sanders, A. E. (2011). Differential impacts of age of acquisition on letter and semantic fluency in Alzheimer’s disease patients and healthy older controls. Quarterly Journal of Experimental Psychology, 64, 2382–2391. doi:10.1080/17470218.2011.596660 Sebastian, N., Martı, M. A., Carreiras, M. F., & Cuetos, F. (2000). LEXESP: L exico informatizado del ~ Barcelona, Spain: Universidad de Barcelona. espanol. Silveri, M. C., Cappa, A., Mariotti, P., & Puopolo, M. (2002). Naming in patients with Alzheimer’s disease: Influence of age of acquisition and categorical effects. Journal of Clinical and Experimental Neuropsychology, 24, 755–764. doi:10.1076/jcen.24.6.755.8407 Taler, V., & Phillips, N. A. (2008). Language performance in Alzheimer’s disease and mild cognitive impairment: A comparative review. Journal of Clinical and Experimental Neuropsychology, 30, 501–556. doi:10.1080/13803390701550128 Tippett, L. J., Meier, S. L., Blackwood, K., & Diaz-Asper, C. (2007). Category-specific deficits in Alzheimer’s disease: Fact or artefact? Cortex, 43, 907–920. doi:10.1016/S0010-9452(08)70690-7 Verma, M., & Howard, R. J. (2012). Semantic memory and language dysfunction in early Alzheimer’s disease: A review. International Journal of Geriatric Psychiatry, 27, 1209–1217. doi:10.1002/ gps.3766 Received 13 July 2014; revised version received 13 May 2015
