Elsevier

Cortex

Volume 47, Issue 9, October 2011, Pages 1052-1062
Cortex

Research report
When two and too don’t go together: A selective phonological deficit sparing number words

https://doi.org/10.1016/j.cortex.2011.03.013Get rights and content

Abstract

We report the case of an Italian speaker (GBC) with classical Wernicke’s aphasia syndrome following a vascular lesion in the left posterior middle temporal region. GBC exhibited a selective phonological deficit in spoken language production (repetition and reading) which affected all word classes irrespective of grammatical class, frequency, and length. GBC’s production of number words, in contrast, was error free. The specific pattern of phonological errors on non-number words allows us to attribute the locus of impairment at the level of phonological form retrieval of a correctly selected lexical entry. These data support the claim that number words are represented and processed differently from other word categories in language production.

Introduction

Several neuropsychological studies have reported selective impairment or sparing of numerical concepts and number words. At the semantic level, double dissociations between numerical concepts and number words have been documented (Cipolotti and Butterworth, 1995, Cipolotti et al., 1991, Cappelletti et al., 2001, Cappelletti et al., 2002, McCarthy and Warrington, 1990, Dehaene and Cohen, 1997; Thioux et al., 1998, Zamarian et al., 2006). The patient described by Cipolotti et al. (1991), for example, was proficient in semantic tasks involving non-numerical material, but was unable to perform numerical reasoning tasks with numbers above four. In contrast, Cappelletti et al. (2001; see also Cappelletti et al., 2005) reported a case of a patient affected by semantic dementia who exhibited profound semantic deficits across different cognitive domains but whose numerical knowledge was intact. Likewise, Zamarian et al. (2006) reported the case of a patient who was markedly impaired in tasks tapping non-numerical semantic knowledge, but whose performance was at ceiling in nearly all numerical tasks. As shown by reaction times, the patient’s processing of arithmetical knowledge was fully preserved and indistinguishable from that of healthy subjects.

The existence of double dissociations between number concepts and other concepts has been taken to suggest that dedicated neural circuitries may be involved in storing and processing numerical cognition (Dehaene and Cohen, 1995, Dehaene and Cohen, 1997, Butterworth et al., 2001, McCarthy and Warrington, 1990, among others). This has been corroborated more recently by neuroimaging studies showing localized activation patterns specific to number processing tasks (see Cantlon et al., 2006, Dehaene and Cohen, 1995, Dehaene et al., 1999, Dehaene et al., 2003, Cappelletti et al., 2007, Thioux et al., 2002 among others).

In contrast to the large body of studies indicating a semantic dissociation between number and non-number concepts, fewer studies have found dissociations between number and non-number words at the post-semantic word production level, in both spoken and written modalities (Basso and Beschin, 2000, Bachoud-Lévi and Dupoux, 2003; Cohen et al., 1997, Geschwind, 1965, Delazer et al., 2002, Marangolo et al., 2004, Marangolo et al., 2005, Dotan and Friedmann, 2010). Cohen et al. (1997) and Bachoud-Lévi and Dupoux (2003) reported on patients who, in the spoken modality with picture naming tasks, made phonological errors on all word categories except for number words. The opposite pattern is reported in Marangolo et al. (2005) who described the case of a patient who made errors only on number words, in the spoken but not in the written modality. In the written modality, Basso and Beschin (2000) reported the case of a patient whose written production of Arabic numerals was impaired, but whose spelling and knowledge of the letters of the alphabet was only minimally affected. This contrasts with Delazer et al. (2002), who reported the case of a patient with a written production deficit that affected the letters of the alphabet but not Arabic numerals.

Dissociations between number words and non-number words may indicate that the language system represents and processes number words separately. However, the level of the language production system where the distinction occurs is unclear. We frame our analysis within a broad consensus framework for language production, as developed by Levelt and colleagues (Levelt, 1989, Levelt et al., 1999; see also Bock and Levelt, 1994, and Ferreira and Slevc, 2007, for a recent review). In this model, spoken word production (e.g., producing the word /kæt/ to describe the picture of a cat) involves the retrieval of two different types of lexical representations: lemmas specified for syntactic–semantic properties, and modality specific lexemes (word forms) specified for morpho-phonological properties (but see Caramazza, 1997, for a model without the lemma/lexeme distinction). In the process of spoken production lexemes are further “unpacked” yielding output phonological representations that then feed into articulatory processes and representations (articulation). The exact nature of these output phonological representations and the amount of phonological and phonetic detail they contain is subject to theoretical debate. For example, some theories (Goldrick and Rapp, 2007) further distinguish between lexical-phonological representations that encode the arbitrary aspects of word forms (e.g., the concept “domestic feline” is arbitrarily associated with the sequence /kæt/ in English and /gatto/ in Italian) and post-lexical phonological representations that encode predictable information (“domestic feline” is phonetically /khæt/ in English). It is not the goal of the current study to distinguish between these more fine-grained levels at the spoken output phonological level prior to articulation, and in our analyses we will keep the broader divisions of the consensus framework for word production as our reference. Within this broad framework, dissociations between the production of number words and non-number words may occur at distinct post conceptual-semantic levels: the level of lemma retrieval, where the syntactic properties of words are specified, or more peripheral output levels of phonological encoding (the lexeme, or the subsequent mapping from the lexeme to lexical and post-lexical representations prior to articulation).

In order to localize a category-specific word production deficit within this functional architecture, an analysis of the type of word production errors is necessary. For example, in order to conclude that a deficit occurs at the level of phonological encoding, one must ascertain that the errors are purely phonological in nature and are not the result of selecting a non-target lemma. This might be difficult to ascertain in the case of aphasic jargon, especially if the errors span over more than one phonological segment or syllable (see Cohen et al., 1997). The number of representational levels and the types of processes also depend on the task. Whereas picture description involves conceptual preparation and lemma selection in addition to lexeme retrieval, repetition and reading tasks may bypass conceptual preparation and lemma selection by recruiting either lexical or sub-lexical acoustic-phonological or graphemic-phonological conversion routes.

In this study, we present data from an individual who exhibits a dissociation in reading aloud between impaired non-number words and spared number words. The patient is similar to the two previously reported ones in the literature (Cohen et al., 1997; Bachoud-Lévi and Dupoux, 2003). There are, however, two crucial differences with the previously reported cases. First the dissociation manifests itself in reading; second, the patient’s error pattern in reading is unusually clear and indisputably phonological because it overwhelmingly affects only one segment, namely vowels. A detailed report on the dissociation between vowels and consonants in this patient’s production was previously reported in Semenza et al. (2007).

Before reporting the case, we briefly review the two most relevant studies indicating the possibility of production deficits sparing the lemma level and involving the phonological level for all word categories except number words.

Cohen et al. (1997) were the first to report in detail a dissociation between non-number words and number words in the production of a Wernicke’s aphasic patient and to locate the deficit post-lexically, at the level of phonological form retrieval. With the exception of number words, on which the patient rarely made errors, the patient exhibited neologistic language production and phonological errors across all word categories in all production tasks, including repetition and reading. His neologistic jargon was characterized by phoneme substitutions (of both consonants and vowels) with preservation of the overall number of syllables of the target in 80% of neologisms. Of these, 54% also preserved syllable structure. The authors did not find differences in error rates for grammatical class, imageability, or an effect of lexical frequency above and beyond length. The exception was number words, on which errors were infrequent, and when present, were mostly lexical (e.g., 250 -> “four-hundred-and-sixty”) or syntactic (e.g., 74 -> seven-hundred-and-forty). The patient made very few phonological errors on number words: only six errors out of 470 number word trials were phonological, while 630 errors out of 886 non-number word trials were phonological. On the basis of the observed sparing of number words from errors, Cohen et al. (1997: 1030) concluded that: “(the) category-specific sparing (of number words) suggests that the cerebral lexicon has a categorical organization down to the phonological level”.

The evidence that Cohen et al. (1997) used to argue for a phonological locus of the dissociation between number and non-number words, however, was somewhat indirect. The main piece of evidence they used was that errors on non-number word targets often matched the overall phonological structure of the target (i.e., number of syllables and syllable structure). Although the specification of syllabic information pertains to the phonological level of encoding and is thus an indication that the patient’s deficit was in retrieving phonological representations, it is not entirely possible to determine whether the patient had selected the appropriate lemma, or whether another lemma with the same number of syllables and syllable structure had been selected.

The second relevant case is reported in Bachoud-Lévi and Dupoux (2003). The patient, GBI, exhibited impaired word production in picture naming tasks sparing numerals, days of the week and months. The patient had very good reading and word repetition.

Unlike the patient in Cohen et al. (1997), this patient’s production was affected by grammatical class: abstract nouns and verbs were significantly less affected than concrete nouns. GBI had spared conceptual knowledge and could correctly assign gender to 95% of the items he could not name, indicating intact lexical access to lemma information. In order to localize GBI’s deficit, the authors examined the naming error patterns. In 70% of the productions, GBI preserved the syllable length of the target. More importantly, the errors were overwhelmingly phonological in nature, consisting of phonemic paraphasias and non-words that often resembled the target in metrical properties (e.g., number of syllables). These phonological errors constituted 86% of the errors, and of these, 78% preserved the number of syllables of the target.

Another indirect piece of evidence suggesting that GBI’s deficit concerned word form retrieval is that variables such as frequency and length affected naming accuracy. These variables have been argued to influence phonological encoding (Jescheniak and Levelt, 1994; but see Dell, 1990 and Navarrete et al., 2006 for evidence that frequency may operate at other levels, including lexical retrieval, and Goldrick and Rapp, 2007 for evidence that lexical frequency operates on what they call lexical-phonological representations, but not post-lexical ones). We will return to the issue of frequency in the General discussion.

In the case we report here, two aspects of the data strongly suggest that the localization of the dissociation is at the level of phonological form encoding: first, given that the task was a reading one, the target word was always known, making lexical miss-selection less likely; second, the errors on non-number words were errors on individual segments, while the rest of the segmental content of the target word was intact. For example, for the target word feroci (/feroʧi/ “fierce”, masculine plural, CV.CV.CV structure) the patient produced the non-word ∗ferici, substituting the fourth segment, i.e., the vowel /o/ with the vowel /i/. The fact that both the syllabic structure and most of the segmental content of the target word are preserved (CV.CV.CV and /f/ /e/ /r/ /ʧ/ /i/), provides very strong evidence that the patient has retrieved the lemma and much of the phonological content of the corresponding lexeme. In sum, the data from this patient may allow a more precise functional localization than previously reported cases and support the proposal that number words are produced differently from other words, down to the phonological level of lexeme retrieval of a selected lemma, or potentially even more “downstream”, in the mapping from the lexeme to the subsequent level of phonological assembly, prior to articulation. In models that distinguish between lexical-phonological and post-lexical phonological representations, such as Goldrick and Rapp (2007), this could even be at the post-lexical-phonological level. We discuss this latter possibility in the General discussion.

Section snippets

Case history

GBC was a 62-year-old businessman, with a long history of vascular problems. In February 1998, he suffered from an ischemic lesion in the left posterior temporal area, resulting in fluent aphasia. An old ischemic lesion in the left occipital lobe was also found on that occasion. In May 1998, he underwent a bypass between the superficial temporal artery and the middle cerebral artery. The procedure was complicated a few days after the intervention by bleeding in the left temporo-parietal area.

Experimental investigation

This investigation was conducted from September 2007 to September 2008. Relative to the first testing period (Semenza et al., 2007) the patient’s symptoms had much improved.

In the experimental reading aloud tasks reported below, he did not make lexical errors.

We report GBC’s reading performance on non-number words belonging to different grammatical categories and varying in frequency and length. We examine the effects of grammatical class, frequency and length on GBC’s errors to ascertain

General discussion

GBC’s performance in reading aloud number and non-number words is striking in two respects. On non-number words, GBC made more phonological selection errors on vowels than on consonants. On number words, in contrast, GBC made no phonological errors.

Semenza et al. (2007) proposed that GBC’s deficit supports a “slot and filler” architecture of word production (e.g., Dell, 1988, Levelt, 1992) for non-number words in which lexemes are further broken down into separable representations: structural

References (50)

  • A.S. Meyer

    Investigation of phonological encoding through speech error analysis. Achievements, limitations, and alternatives

    Cognition

    (1992)
  • C. Romani et al.

    Patterns of phonological errors as a function of phonological versus an articulatory locus of impairment

    Cortex

    (2002)
  • C. Semenza et al.

    A dedicated neural mechanism for vowel selection: A case of relative vowel deficit sparing the number lexicon

    Neuropsychologia

    (2007)
  • C.A. Sevald et al.

    Syllable structure in speech production: Are syllables chunks or schemas?

    Journal of Memory and Language

    (1995)
  • R. Treiman

    The structure of spoken syllables: Evidence from novel word games

    Cognition

    (1983)
  • L. Zamarian et al.

    Knowing 7×8, but not the meaning of ‘elephant’: Dissociation between numerical and non-numerical semantic knowledge

    Neuropsychologia

    (2006)
  • A.C. Bachoud-Lévi et al.

    An influence of syntactic and semantic variables on word form retrieval

    Cognitive Neuropsychology

    (2003)
  • A. Basso et al.

    Number transcoding and number word spelling in a left-brain-damaged non-aphasic acalculic patient

    Neurocase

    (2000)
  • P.M. Bertinetto et al.

    Corpus e Lessico di Frequenza dell’Italiano Scritto (CoLFIS)

  • S.E. Blumstein

    Deficits of speech production and speech perception in aphasia

  • J.K. Bock et al.

    Language production: Grammatical encoding

  • B. Butterworth et al.

    Category specificity in reading and writing: The case of number words

    Nature Neuroscience

    (2001)
  • J.F. Cantlon et al.

    Functional imaging of numerical processing in adults and 4-year-old children

    PLoS Biology

    (2006)
  • M. Cappelletti et al.

    Why semantic dementia drives you to the dogs (but not to the horses): A theoretical account

    Cognitive Neuropsychology

    (2002)
  • M. Cappelletti et al.

    Dissociations in numerical abilities revealed by progressive cognitive decline in a patient with semantic dementia

    Cognitive Neuropsychology

    (2005)
  • Cited by (16)

    • Numbers, calculation and acalculia

      2021, Encyclopedia of Behavioral Neuroscience: Second Edition
    • Common neural circuit for semantic-based articulation of numbers and words: A case study of a patient with Broca's aphasia

      2021, Journal of Neurolinguistics
      Citation Excerpt :

      Studies have shown a double dissociation between the neural correlates for words and numbers (Bencini et al., 2011; Cohen, Verstichel, & Dehaene, 1997; Dotan & Friedmann, 2015; Piras & Marangolo, 2009). Numerous studies have reported patients with injuries to the temporal cortex who exhibited poor performance in processing words, but who processed numbers normally (Bencini et al., 2011; Butterworth, Cappelletti, & Kopelman, 2001; M.; Cappelletti, Butterworth, & Kopelman, 2001; Marinella; Cappelletti, Butterworth, & Kopelman, 2006; M.; Cappelletti, Kopelman, & Butterworth, 2002; M.; Cappelletti, Kopelman, Morton, & Butterworth, 2005; Cheng et al., 2013; Cohen et al., 1997). For example, Cappelletti et al. (2005) reported a patient with semantic dementia who displayed impaired semantic processing of words (e.g., semantic fluency, category naming, and word classification), but relatively normal numerical processing (e.g., number seriation, number composition, and calculation).

    • One can be some but some cannot be one: ERP correlates of numerosity incongruence are different for singular and plural

      2019, Cortex
      Citation Excerpt :

      The great majority of studies investigating the relationship between numerical knowledge and its encoding into language have taken into consideration the lexical level, mainly focused on quantifiers and number words (e.g., Butterworth et al., 1999; Carey, 2004; Clark & Grossman, 2007; Gelman & Gallistel, 2004; Gordon, 2004; Lipton & Spelke, 2003; Ochtrup et al., 2013; Rath et al., 2015; Salillas, Barraza, & Carreiras, 2015; Troiani, Peelle, Clark, & Grossman, 2009). It has been shown that speakers of languages without number words do master non-verbal numerical skills (e.g., Butterworth, Reeve, Reynolds, & Lloyd, 2008; Pica, Lemer, Izard, & Dehaene, 2004), being these latter independent from the verbal ones, and that number words are dissociated from other lexical categories (Bencini et al., 2011; Semenza et al., 2007). However, the information about numerosity can be expressed in language without relying on number words by means of Number morphology, which systematically encodes it into different signs (e.g., cat vs cats in English).

    • Separate mechanisms for number reading and word reading: Evidence from selective impairments

      2019, Cortex
      Citation Excerpt :

      Reading numbers aloud involves parsing a digit string and translating it into a sequence of number words. The visual analysis of digit strings and the verbal production of number words are implemented by separate cognitive mechanisms, as indicated by several neuropsychological case studies (Bencini et al., 2011; Benson & Denckla, 1969; Cohen & Dehaene, 1995; Cohen, Verstichel, & Dehaene, 1997; Delazer & Bartha, 2001; Dotan & Friedmann, 2015; Dotan, Friedmann, & Dehaene, 2014; Marangolo, Piras, & Fias, 2005; Marangolo et al., 2004; McCloskey, Sokol, & Goodman, 1986; Noël & Seron, 1993) and brain imaging studies (Dehaene & Cohen, 1995; Dehaene et al., 2003). Turning to finer-grained details, we describe a cognitive model that we developed (Dotan & Friedmann, 2018), which merges and extends previous models (Cohen & Dehaene, 1991; McCloskey, 1992; McCloskey et al., 1986).

    • Steps towards understanding the phonological output buffer and its role in the production of numbers, morphemes, and function words

      2015, Cortex
      Citation Excerpt :

      Several studies reported phonological errors that appeared in non-number words but not in numbers. Bencini et al. (2011) and Semenza et al. (2007) investigated GBC, a man with Wernicke's aphasia who produced numbers flawlessly but made many phonological errors in word production (only in vowels). Patient LT (Shallice, Rumiati, & Zadini, 2000) had an impaired phonological output buffer and made many phonological errors in word production.

    View all citing articles on Scopus
    View full text