Disentangling developmental prosopagnosia: A scoping review of terms, tools and topics

The goal of this preregistered scoping review is to create an overview of the research on developmental prosopagnosia (DP). Through analysis of all empirical studies of DP in adults, we investigate 1) how DP is conceptualized and deﬁned, 2) how individuals are classiﬁed with DP and 3) which aspects of DP are investigated in the literature. We reviewed 224 peer-reviewed studies of DP. Our analysis of the literature reveals that while DP is predominantly deﬁned as a lifelong face recognition impairment in the absence of acquired brain injury and intellectual/cognitive problems, there is far from consensus on the speciﬁcs of the deﬁnition with some studies emphasizing e.g., deﬁcits in face perception, discrimination and/or matching as core characteristics of DP. These differences in DP deﬁnitions is further reﬂected in the vast heterogeneity in classiﬁcation procedures. Only about half of the included studies explicitly state how they classify individuals with DP, and these studies adopt 40 different assessment tools. The two most frequently studied aspects of DP are the role of holistic processing and the speciﬁcity of face processing, and alongside a substantial body of neuroimaging studies of DP, this paints a picture of a research ﬁeld whose scientiﬁc interests and aims are rooted in cognitive neuropsychology and neuroscience. We argue that these roots e alongside the heterogeneity in


Introduction
Prosopagnosia is an impairment of the human ability to recognize other people by their face (Starrfelt & Barton, 2022).Individuals with prosopagnosia struggle to recognize friends, family, colleagues and e in some cases e even their own reflection in the mirror.For many this has negative psychosocial consequences with an impact on quality of life (Yardley et al., 2008).A distinction can be made between two types of prosopagnosia; prosopagnosia induced by brain injury such as stroke or traumatic brain injury (acquired prosopagnosia) and prosopagnosia believed to be present from birth (Cook & Biotti, 2016).This latter type has been referred to by several different names in the literature: Congenital Prosopagnosia (CP), Hereditary Prosopagnosia (HP) and Developmental Prosopagnosia (DP).CP refers to innate prosopagnosia (Behrmann & Avidan, 2005), while HP emphasizes the genetic aspect of prosopagnosia (Kennerknecht et al., 2006).As such, both these concepts include prosopagnosia caused by prenatal factors.In this context, however, we will use the third term, DP, which, in addition to subjects with CP/ HP, includes cases where the cause of prosopagnosia is not known, which is the majority of published cases.Individuals with DP have life-long difficulty in recognizing or learning to recognize faces.Barton and Corrow (2016) referred to this definition of DP as deceptively simple (p.119), and there is indeed far from consensus on the specifics of how to define the phenomenon at hand.This dissent ranges from defining DP as the failure to develop a face recognition system (de Gelder & Stekelenburg, 2005) to defining it as a condition in which face perception is impaired in the absence of an obvious brain damage and given intact sensory and intellectual functions (Tanzer et al., 2016).These differences in definitions constitute a problem for the research field, because it makes it unclear whether different research groups are dealing with the same phenomenon, and thus whether findings are transferable to DP populations in other contexts.
The differences in DP definitions and the fact that several different concepts are used to denote what is assumed to be the same phenomenon, suggests disagreement in the field about what DP is.Scratching the surface reveals not only dissent about whether the condition is best conceptualized as developmental or innate, but a fundamental discussion about the P in DP, that is, what are the core characteristics of this type of prosopagnosia, and should it even be called prosopagnosia in the first place.This discussion unfolded in a special issue of Cognitive Neuropsychology on DP from 2018, spearheaded by a review by Geskin and Behrmann (2018) examining the association or dissociation between face and non-face object processing in DP, that is, whether the impairment in DP is face specific.The review sparked a vigorous debate on e among other issues e how to classify cases of object agnosia (Garrido et al., 2018), how to meaningfully compare deficits in non-face object processing to deficits in face processing (e.g., Campbell & Tanaka, 2018;Gerlach et al., 2018), how to approach these issues methodologically (de Gelder & Van den Stock, 2018;Eimer, 2018;Ramon, 2018;Starrfelt & Robotham, 2018), and whether the term prosopagnosia should even be used in non-acquired cases, as discussed by Rossion (2018a) who proposed the competing term prosopdysgnosia for the lifelong variant.Rossion (2018b) has put this critique forth in the context of acquired prosopagnosia research also, suggesting that the term prosopagnosia should be reserved for neuropsychological cases of recognition impairments specific to faces, more strictly in line with Bodamer's (1947) original definition.It should be noted, though, that the specificity of the face recognition impairment in Bodamer's original cases has also been a subject of discussion (Ellis & Florence, 1990;Gainotti, 2010).Together, this debate encircles whether non-face processing deficits are e and indeed even meaningfully can be e a part of the DP condition, and so the definition of prosopagnosia and even more so of developmental prosopagnosia, remains unsettled.Therefore, the first aim of this review is to map all conceptualizations and definitions of DP in the empirical literature.A conceptualization is here understood as the term used to denote prosopagnosia (i.e., DP or CP), while a definition is the concrete description of the psychological phenomenon (e.g., lifelong face recognition deficit with no brain injury).
The heterogeneity in DP conceptualization and definition is closely tied to the issue of how to classify individuals with DP.The instruments used to detect a phenomenon set the epistemological scene; what DP is e and can be e is delimited by the diagnostic tools and criteria used to classify it.So far, no universal criteria for classifying a subject with DP have been established, and the means of classification vary substantially, with different research groups adopting different classification criteria and a wide range of performance tests, subjective measures and diagnostic interviews (Barton & Corrow, 2016;Bate & Tree, 2017;DeGutis et al., 2023).Some researchers use rather strict classification criteria such as scores below 2 standard deviations from control mean on two or more face processing tests such as the Cambridge Face Memory Test (CFMT) (Avidan et al., 2011;Duchaine & Nakayama, 2006).Others recommend relying only on self-report measures such as the 20-Item Prosopagnosia Index (PI20, Shah, Gaule, Sowden, et al., 2015) when classifying DP, as some individuals who report severe impairment in face recognition are able to perform within control range on objective face processing tests (Burns et al., 2022).Others still have argued that DP classification should include both objective and subjective assessment tools (Arizpe et al., 2019;Barton & Corrow, 2016).Because of these differences in how individuals are classified with DP, the second aim of our review is to map all the assessment tools actively used for DP classification in the literature.
Upon establishing how DP is defined, denoted and diagnosed across the DP research field, we finally examine which aspects of DP are investigated in the literature.The rationale of this exploration is, that if there is a lack of consensus on what DP is and how to assign individuals to the DP category, these issues must be reflected in the DP research questions.As an example, the previously discussed issue of whether nonface object processing impairment is a part of DP is presumably reflected in a large number of studies investigating object processing and the specificity of the face processing impairment in DP.As such, the idea of this third part of our review is to examine which questions are asked in the DP literature rather than which answers are given.
Based on these general problems in the current state of the DP research field, our scoping review asks the following three questions: 1) How is DP defined and conceptualized in the literature?2) How are individuals classified as having DP in the literature? 3) Which research questions about DP are asked in the literature?
To answer these questions, we perform a systematic, exhaustive search of all empirical studies of DP and extract information from them based on the PRISMA-ScR guidelines for conducting a scoping review (Tricco et al., 2018).The review protocol was preregistered in June 2022 and is available c o r t e x 1 7 6 ( 2 0 2 4 ) 1 6 1 e1 9 3 here: https://osf.io/k9er2.The protocol was developed based on the guidelines for scoping review protocols proposed by Peters et al. (2022).

Search strategy
A systematic literature search was conducted in the following databases: PubMed, Scopus, Embase, Web of Science and PsycInfo.
In each database the search consisted of a combination of free text search and, when possible, a search using terms from a controlled vocabulary.This latter search type was only possible to include in the PubMed search, as it is the only one of the selected databases with a controlled vocabulary containing a topic specifically for the variant of prosopagnosia with no brain damage (i.e., the MeSH term 'Prosopagnosia/ congenital').In e.g., PsycInfo the thesaurus only contains the broad topic 'Prosopagnosia', which would create too many irrelevant search results, as it includes all indexed articles on acquired prosopagnosia in addition to those on DP.
Thus, the free text search was the backbone of the systematic search.This aligns well with the fact that the review is tied to the concept of DP and its synonyms.We can assume that all records relevant for the review will contain these words e otherwise, they would not be relevant.The free text search consisted of 'developmental prosopagnosia' and its synonyms separated by the Boolean OR separator.To capture all relevant records of DP, the search string contained synonyms to DP that are infrequently used (i.e., 'childhood prosopagnosia' and 'prosopdysgnosia').
Search results were limited to publications from the year span 1976e2022 and limited to records written in English.The complete search string for each individual database is listed in Table 1.

Eligibility criteria
Interestingly, the majority of studies of DP include adult participants, even if it is considered a developmental disorder, and indeed, measuring face recognition and classifying DP in children remains challenging (Bennetts et al., 2017;Dalrymple, Garrido, & Duchaine, 2014).Thus, we included only peer-reviewed articles that examine adult human participants with DP in accordance with the definition of DP given above (cf.Supplementary Table S4 for a list of studies investigating DP in children).Studies that investigated only children with DP were excluded in order to better disentangle the studied phenomenon from other developmental conditions.As mentioned, different classification criteria for DP are used across the literature, so we included all accounts of prosopagnosia without known brain damage.As such, inclusion of studies in the review is tied to the concept of DP rather than a set of criteria for classifying DP.In other words, any empirical article that claims to investigate subjects with DP (or any of its synonymous concepts) was eligible for inclusion.The empirical material could be of any kind; quantitative, qualitative, collected in a lab setting, online or in the field, but only original, peer-reviewed accounts of the data were considered.Data that was merely referenced from other sources was not included.Other reviews, conference abstracts and theoretical commentary articles on DP were also not included.
As the first known account of developmental prosopagnosia is from 1976 (McConachie, 1976), it was deemed appropriate to limit the search of literature to the years 1976e2022.Earlier accounts of DP may exist, but to keep the phenomenon within reasonable conceptual boundaries, we limited the search to literature published after the first registered case of developmental prosopagnosia was described.The upper year limit of 2022 was chosen in accordance with the preregistered review protocol.For an overview of more recent publications, the reader is referred to section 3.4.

Selection process
The ("developmental prosopagnosia" OR "congenital prosopagnosia" OR "childhood prosopagnosia" OR "hereditary prosopagnosia" OR "prosopdysgnosia") AND DT 1976e2022 c o r t e x 1 7 6 ( 2 0 2 4 ) 1 6 1 e1 9 3 To qualify the assumption that we made an exhaustive search and an appropriate selection, we crosschecked our selected records with the sources included in previous reviews of the field (Geskin & Behrmann, 2018;Kress & Daum, 2003a).The inclusion and exclusion process is shown in Fig. 1.

2.4.
Data synthesis To answer our three research questions, a wide range of information was extracted from each included record and registered in a spreadsheet, available here: https://osf.io/pmnhj (see also Supplementary Table S1).The extracted information is outlined below and the preregistered review protocol can be referred to for a full description.In the data synthesis process all included records were treated equally in the sense that the extracted information was not weighted based on sample size or other metrics.As such, a single case study (e.g., Bentin et al., 1999, N DP ¼ 1) contributed with the same amount of information to all our analyses as did a study of a larger sample of DP individuals (e.g., Little et al., 2022, N DP ¼ 101).Although a weighting of studies based on sample size or a case-by-case analysis akin to the approach by Geskin and Behrmann ( 2018) is well suited to synthesize knowledge about a given research topic (e.g., object recognition), a drawback of such an approach is that the same (larger) samples of DP individuals appear in more than one publication (e.g., Liu et al., 2021;Zhao et al., 2016) which would for our purposes result in an over approximation of the significance of those samples on the field.Further, sample sizes have organically grown over time as the access to online participant recruitment and testing alongside public awareness of DP has increased, and accounting for sample size in our analyses might underestimate the relative impact of the early, pioneering, small-scale DP studies.Finally, some research topics are arguably better suited for smaller samples and case studies such as in-depth investigations of the psychosocial consequences of DP (e.g., Diaz, 2008), while larger samples are required for investigations of e.g., symptom heterogeneity (Bate, Bennetts, Gregory, et al., 2019).As such, we have not emphasized sample size variability in our review of the field (e.g., by weighting the information in Figs.2e4 based on sample sizes), however the reader may be referred to Supplementary Fig. S1 for an overview of how sample sizes have changed over time.

Conceptualization and definition
For each record, the concept used to denote prosopagnosia (DP, CP or HPA) was registered.Further, text sections from each record containing definitions of DP were manually coded into definition 'building blocks' by the first author, and all codings were affirmed by the last author.For instance, Palermo et al (2011b) defines DP as follows: People with congenital prosopagnosia (CP; also referred to as developmental prosopagnosia) have severe, life-long deficits recognising the identity of familiar people from their faces despite intact low-level vision and general cognitive abilities.This definition was coded into consisting of the following four building blocks: Lifelong þ Face identification impairment þ Absence of visual or other sensory problems þ Absence of intellectual/cognitive problems.The process of coding text into definition building blocks was accompanied by a certain amount of reduction and merging definitions that vary slightly in semantics.For instance, the terms impairment and deficit were coded interchangeably here, and all for all relevant building blocks, impairment was used, meaning that studies that define DP as e.g., a face recognition deficit were coded as face recognition impairment.The guiding principle in the coding process was to strike a balance between highlighting the heterogeneity of DP definitions while keeping at a reasonably low and surmountable number of definition building blocks.As such, new building blocks were added when studies were encountered that provided definitions that differed in conceptually important ways from already coded studies.As an example, a distinction was made between face recognition impairment and face identification impairment building blocks, because it is conceptually significant whether face recognition (knowing that you have seen someone before) or face identification ability (being able to identify specific individuals based on their face) is considered the defining feature of DP.

Classification of DP
The primary focus when mapping how individuals are classified with DP was to look at which diagnostic measures are adopted in the classification of DP.First, however, we made a crude distinction between studies that explicitly described how individuals were classified, and studies that did not contain such descriptions.This distinction was made based on a rather strict criterion: Studies were only considered to have descriptions of DP classification if they listed explicit criteria for diagnosing subjects with DP.Many studies (especially some of the early, pioneering studies of DP) describe thoroughly which tests they use to assess the subjects' face processing problems.
They also report the scores on these tests and (sometimes) how much they deviate from control scores.But they rarely list how many of these scores must deviate and by how much in order to classify a subject as a DP.Such assumptions may indeed exist for the relevant studies.The authors may have had criteria for diagnosing DP that are not listed in the papers themselves.For instance Nunn et al. (2001) thoroughly describe a neuropsychological examination of an individual with DP using a total of nine different tests of face processing, but the study does not explicitly specify any diagnostic criteria.In other cases, the presence of criteria is more unclear, so a certain amount of interpretation of the text was necessary.For instance, in the case study of a horse expert by Weiss et al. (2016), the DP subject is described as having z-scores of À2.16 and À2.45 compared to control means on the CFMT and an FFT, respectively.Additionally, the authors define 'the threshold for abnormal face perception in the specified tests was defined as 2 SDs below average similarly to other papers' (Weiss et al., 2016, p. 67).Does this mean that the DP subject would not have been classified as a DP, if she had not fallen below 2 SD's on both tests?Would an abnormal (>2 SD's) performance on only one of the diagnostic tests be sufficient for a DP classification?As such, it is a general assumption of this review that the criteria for category classification of DP is determinant for what DP is regarded to be.More or less liberal diagnostic criteria come with a definitional power e if, for instance, a study explicitly uses a face perception test as a diagnostic measure, face perception impairment becomes an integral part of what DP is regarded to be.Additionally, the prevalence of DP depends heavily on the diagnostic criteria adopted (cf.Barton & Corrow, 2016;DeGutis et al., 2023).Thus, any variability in how DP is classified across the literature is worth scrutinizing.For all included records, the measures involved in the classification process were registered.For the studies that explicitly described how individuals were classified as having DP, these diagnostic tools were examined in depth (see Results).

Research questions in the DP literature
To investigate which research questions are most frequently asked in the DP literature, all included studies were grouped based on which specific aspect(s) of DP, they studied.The aim of this was to systematically examine and illustrate which aspects of DP are most frequently investigated, in order to point at potential problems and gaps in the existing literature.

Results and discussion
A total of 224 studies were deemed eligible for review.Characteristics of all studies can be explored in the Supplementary Table S1.Of the 224 records included here, 197 papers stated clear definitions of the phenomenon DP.Coding these 197 definitions into 'building blocks' yielded a total of 18 different blocks.4 of these (e.g.Lifelong and Hereditary basis) pertained to the 'developmental' part of DP, while 14 of the blocks can be thought of as positive/present and negative/absent properties of the 'prosopagnosia' part of DP (e.g., face recognition impairment and absence of acquired brain injury).The result of this coding process is visualized in Fig. 2.

Conceptualization and definition of DP
As the figure indicates, the predominant definition of DP is that it is a lifelong face recognition impairment in the absence of acquired brain injury and in the absence of intellectual/cognitive problems.Other buildings blocks appear more sporadically in the literature.In 11 studies, DP is defined as having a component of face perception impairment.Intact face perception is definitely a condition for successful face recognition, and as such the face learning/recognition/identification impairment blocks may be thought of as nested within the face perception impairment block.It is interesting, however, that face perception problems (and in 12 studies face matching/ discrimination) is an explicit part of the definition of DP in only a minority of studies.This implies that DP participants in most studies do not necessarily have problems with face perception, but in many cases this was not measured.Some studies (n ¼ 20) refer to the broader face processing impairment which may refer to a breakdown of any process necessary for successful face identification.Another conceptual nuance worth highlighting is that 33 definitions of DP contain the building block face identification impairment rather than face recognition impairment.While it is not entirely clear what this particular difference in definition implies in all instances in the literature, we may refer to the distinction made by Rossion (2022).Here, face recognition is defined as the judgment of a previous occurrence of specific identities of faces, while face identification (or what Rossion refers to as 'face identity recognition') is the production of a unique response to a given face according to its individually distinctive characteristics (Rossion, 2022, p. 4).As an example, meeting a colleague at the supermarket may invoke a sense of familiarity or recognition (the former process), but not the ability to remember who they are, their name, or distinctive semantic information about them (the latter process).Whether this distinction is what is reflected in the discrepancy between defining DP as an impairment in face recognition or face identification, is not obvious, but it is at least worth noting that this discrepancy exists.
Of course, the heterogeneity in DP definitions need not reflect a disagreement among researchers about what DP is.Rather, it is likely that different researchers highlight different core features of DP in their introduction to frame the problem they are addressing in their respective paper.However, the lack of consensus on how specifically to define DP suggests that a conceptual gap exists in the research field.This gap is literally represented by the tendency to define DP by its absence of properties (absence of brain injury, absence of visual problems, etc.).The focus on absence may be explained by DP's conceptual origin as an offspring of acquired prosopagnosia.Historical factors may have led to DP being defined mainly by how it differs from acquired prosopagnosia, and e as we will argue later e this is also reflected in the DP field adopting research questions and theoretical assumptions from acquired prosopagnosia research.

DP classification
Of Aside from differing methodologically, the different assessment approaches arguably tap different aspects of the DP phenomenon.For instance, the Famous Faces Tests measure semantic knowledge of celebrity faces (Wilson et al., 1981), the Cambridge Face Memory Test (CFMT) measures the ability to learn small differences in unfamiliar faces (Duchaine & Nakayama, 2006), the Cambridge Face Perception Test measures face perception performance (Duchaine et al., 2007), while the PI20 is a self-report measure of the ability to recognize faces on a day-to-day basis (Shah, Gaule, Sowden, et al., 2015).As such, the use of many different combinations of diagnostic tests is likely to constitute a sampling problem for the DP field.Unless a case can be made that the exact same population of individuals would score abnormally on all diagnostic tools outlined here, different research groups may well be dealing with different psychological phenomena simply as a consequence of adopting different diagnostic measures.This heterogeneity in how DP is assigned to individuals contribute to the loose conceptualization of DP.Further, the use of different classification criteria (e.g., being 1.7 or 2.0 SD's below the mean of a control group on one or more diagnostic tests), which is not directly addressed in this review, exacerbates the conceptual issue constituted by the use of many different diagnostic tools, and has recently been discussed elsewhere (e.g., DeGutis et al., 2023).c o r t e x 1 7 6 ( 2 0 2 4 ) 1 6 1 e1 9 3 Here, our primary aim is to highlight that a diagnostic heterogeneity exists and stress the importance of clearly describing how individuals are classified with DP in a given study.After all, the heterogeneity in DP classification may be an indication of the variability observed in the DP phenomenon itself e for instance, Johnen et al. (2014) found overall poor face recognition ability with varying patterns of impairment in non-face processing within a family where some members met the diagnostic criterion for DP.As such, scrambling for strict diagnostic consensus may not be a goal in itself.However, DP classification in future research would benefit from striving towards criteria that capture the three main conceptual components of DP, i.e., 1) that it is a visual agnosia, understood as the inability to recognize visual stimuli (Gerlach & Robotham, 2021;Tranel & Damasio, 2001), 2) that it is face specific, as the term prosopagnosia implies a problem clearly disproportionate for recognizing faces (Lahiri, 2020) and 3) that the face recognition deficit is neurodevelopmental, i.e., not a result of brain injury.In this context, solely relying on self-report measures for DP classification (cf.Burns et al., 2022) seems inadequate, as people cannot be expected to distinguish between an impairment in recognizing facial identity and e.g., impairments of low-level vision, and further may not be able to judge whether their difficulties are disproportionately larger for faces compared to other visual categories.In addition, some individuals with DP may not be aware of the severity of their impairment, but may assume (as most of us do) that other peoples' face recognition ability is on level with their own.DP classification, then, should ideally rely on a combination of objective and subjective assessment, as others have suggested (Arizpe et al., 2019;Barton & Corrow, 2016).With regards to the specificity of face recognition impairment in DP classification, we recently proposed adopting an approach contrasting face and non-face object recognition performance (Gerlach et al., 2024).If the diagnostic logic inherited from clinical neuropsychology is not applied, one could argue that the term prosopagnosia is not even appropriate for the behavioral profile of DP (see Rossion (2018a) for an elaborate discussion of this point).Finally it should be noted, that in this review we assumed the diagnostic criteria to be a demarcation of what DP is regarded to be, and this assumption may not be applicable to all DP studies.For instance, Dobel et al. (2007) adopted a conservative criterion to ensure that the included patients exhibited severe face recognition difficulties.In such a case, classification criteria are not criteria for diagnosis as such, but rather for symptom severity (see also DeGutis et al., 2023).Increased clarity in describing how DP is diagnosed by individual research groups may help alleviating this uncertainty.

Investigated aspects of DP
The third aim of this review was to examine which questions are asked in the DP literature.At first glance, it is obvious that DP research is firmly rooted within the disciplines of cognitive neuropsychology and neuroscience; the three by far most frequent scientific outlets of the reviewed studies are Neuropsychologia (n ¼ 33), Cortex (n ¼ 29) and Cognitive Neuropsychology (n ¼ 22), and roughly one third (n ¼ 76) of all studies utilize neuroimaging techniques.This tendency can be further explored by examining which aspects of DP are investigated.To this end, the main topics of investigation were derived from each of the included studies, with each study investigating between one and three overall topics.Fig. 4 shows the frequency of each of these topics.Infrequent topics (topics that were investigated in only one study) were collapsed into one category in the figure and included topics such as musical pitch in DP (Corrow et al., 2019b), processing of pareidolic objects (Epihova et al., 2022) and face trustworthiness judgments (Todorov & Duchaine, 2008).These infrequent topics are listed in Supplementary Table S2.45 of the 76 studies utilizing neuroimaging techniques are concerned with mapping the neural correlates of DP by searching for structural (e.g., Garrido et al. (2009)) and functional (e.g., Gerlach et al. (2019)) differences between DP's and controls.A substantial part of this literature (n ¼ 13) is specifically focused on whether individuals with DP exhibit neurotypical N170/ M170 potentials using EEG/MEG, and so this topic received its own category in Fig. 4. The neuroimaging literature on DP was recently thoroughly reviewed by Manippa et al. (2023), and for our purposes the main issue to note is that the controversies related to definition and classification of DP may pose a challenge for determining the neural correlates of the condition.
Two of the most frequently investigated cognitive topics across the DP literature are holistic processing (n ¼ 32, including papers that investigate 'configural' processing) and the issue of the specificity of faces (n ¼ 25).The majority of the 32 papers on holistic processing aim to test the hypothesis that the face recognition impairment in DP is due to a general deficit in the ability to integrate parts into a whole.Holistic processing may be defined as a perceptual phenomenon gluing together the features into a gestalt (Esins et al. (2016), p. 2).The holistic processing hypothesis for faces originates in research on acquired prosopagnosia, where the contested issue is whether the loss of face recognition ability is a category specific deficit for faces (McNeil & Warrington, 1993) or a result of the breakdown of one or more domain general processes necessary for processing faces (Farah, 1990;Gauthier et al., 1999).This issue has also been thoroughly discussed in the Each black bar represents one study and the height of the bar indicate the sample size of the DP group of the respective study.A link between two bars indicate that the two tools are used in one study in combination to classify DP.Tools that exist in more than one edition (e.g., the FFT's) are accompanied by references to the study in which that respective edition was used.The distinction between tests of face perception and face memory was done on the basis of the principles outlined by Robotham and Starrfelt (2018) context of neuroimaging studies, specifically regarding the existence and role of the fusiform face area (FFA, e.g., Gauthier, 2017;Kanwisher, 2017).Duchaine (2000) provided the first investigation of the holistic processing hypothesis in DP by reporting normal configural processing in a 53-year-old individual with DP, interpreting this as evidence against the holistic processing hypothesis, and in favour of a face specific deficit.However, other studies point in the opposite direction (Avidan et al., 2011;Carbon et al., 2007), and the sheer amount of research on holistic processing in DP conducted since Duchaine's 2000 study e paired with the fact that the studies are scattered across publication years e suggests that this has been and remains an unresolved debate.
The related topic of the specificity of the impairment in DP e does it affect only faces or also other visual objects (n ¼ 25) e is directly linked to the question of how DP should be defined, i.e., if non-face object processing impairment or similar should be a definition building block or an exclusion criterion.The papers in this category include studies that investigate the processing of human bodies (e.g., Rivolta et al. (2017), horses (Weiss et al., 2016) and cars (Gray et al., 2019), as well as mixed object classes (e.g., Gerlach et al., 2016).One could argue that even the studies of word processing/reading in DP (e.g., Rubino et al., 2016;Starrfelt et al., 2018) belong in this group, although we have classified these papers separately in Fig. 4. The underlying motivation for examining the question of face specificity is not unrelated to that of the holistic processing hypothesis; if face processing is selectively impaired in DP, it suggests that faces can be dissociated from other perceptual categories.This in turn may be interpreted as reflecting an underlying face specific mechanism or process.The point here is that two of the most frequently studied topics in DP research both tackle a debate that arose within the cognitive neuropsychology of acquired brain injury.Many of these studies also aim to draw conclusions about the cognitive architecture of neurotypical face processing, as is common in cognitive neuropsychological studies of acquired cases.Another example of this is the issue of covert recognition in DP (n ¼ 11), which was also originally studied in cases of acquired prosopagnosia, as it was believed that some prosopagnosics were in fact able to recognize faces without being aware of it (de Haan et al., 1987), and that this enabled inferences about neurotypical face processing.
As such, both the holistic processing hypothesis and the issue of the specificity of faces in DP (and to some extent the issue of covert recognition) are installations of perhaps the most central debate in cognitive neuropsychology, namely the debate about the modularity of the cognitive system (Coltheart, 2004;Plaut, 1995;Shallice, 1988).In relation to face processing, this debate is concerned with the question of whether visual cognition relies on specialized, dedicated modules for different visual categories, e.g., faces and words, or whether domain general processes contribute to visual (re) cognition of several types of objects, although perhaps differentially (Farah, 1990;Behrmann & Plaut, 2014).In this perspective, a holistic processing impairment in DP (as shown in e.g.Avidan et al. (2011)) could be interpreted as evidence in favor of the latter position, on the basis that holistic processing is one of the general-purpose mechanisms responsible for several perceptual tasks (e.g.face processing).The presence of non-face object processing impairments in DP (as reviewed thoroughly in Geskin and Behrmann (2018)) could be interpreted similarly following the logic that the processing of different perceptual categories rely on shared cognitive architecture, rather than separate specialized modules for processing face and non-face objects.This debate in relation to face processing is thus related to a similar discussion within cognitive neuroscience/neuroimaging (Dehaene & Cohen, 2011;Kanwisher, 2017;Kanwisher et al., 1997;McCandliss et al., 2003) and the neuropsychology of acquired agnosias (Behrmann & Plaut, 2014;Kleinschmidt & Cohen, 2006;Rice et al., 2021;Rossion, 2022).Indeed, many of the research questions posed in the DP literature aim to draw conclusions or make theoretical inferences about the cognitive architecture of (neurotypical) face processing, as one would in cognitive neuropsychological studies of acquired deficits.However, this might not be straightforward.When prosopagnosia arises from acquired brain injury, we can be fairly certain that face recognition has changed from unimpaired to impaired based on the lesion, and thus e by the logic of cognitive neuropsychology (Coltheart, 2004;Shallice, 1988) e we may draw inferences about what the system must have been like before the injury.In developmental disorders, this logic does not necessarily apply (Bishop, 1997;D'Souza & Karmiloff-Smith's 2011).When we study DP, we are investigating a cognitive system that likely had a different starting point and developmental trajectory, and thus might operate in ways very different from the neurotypical face processing system.Thus, it is not clear whether DP research can function as a tool to make theoretical claims about the cognitive architecture of (neurotypical) face processing (Rossion, 2018a;Starrfelt & Robotham, 2018), at least not without additional assumptions.This can perhaps explain why many of the issues inherited from the research on acquired prosopagnosia (i.e., holistic processing, the specificity of face processing, and covert recognition) remain contested in DP research.
Another striking characteristic of the DP literature, as Fig. 4 shows, is that 18 studies investigated an infrequent topic, indicating that it is quite common to research aspects of DP that no one else does.As DP is a relatively new research field, it is not surprising e and not necessarily disadvantageous e that many different aspects of the phenomenon are explored.However, a sign of bifurcation in the literature is clear: on one hand, the DP field is occupied with research questions originating in the acquired prosopagnosia literature based on a framework adopted from cognitive neuropsychology and neuroscience.On the other hand, the field investigates an abundance of niche aspects of DP that have never quite "caught on" as research topics.It is striking that aspects of what DP is such as its hereditability (n ¼ 4), its prevalence (n ¼ 5) and of own and other face perception in DP (n ¼ 3 and 5, respectively) are so relatively rare.Additionally, only a few studies have aimed to develop therapeutic approaches to DP (5 training/rehabilitation studies and 2 greeble training studies), suggesting that alleviating face recognition problems for individuals with DP has not been a priority for the field.It should however be noted, that quite a few (n ¼ 17) studies investigate symptom heterogeneity in DP, suggesting that some research is motivated by characterizing what DP is rather than using it as a tool to describe the neurotypical face processing system.

Recent publications
As the protocol for this review was preregistered in June 2022, only records up until then were included in the present analyses.However, a brief look at the DP literature published after June 2022 is warranted in order to ascertain whether the most recent literature digresses from the literature we reviewed in depth.A follow-up literature search in the five databases (PubMed, Scopus, Embase, Web of Science and PsycInfo) conducted on 23-11-2023 yielded 16 relevant records of empirical studies of DP published since the initial June 2022 search.Relevant records were identified by screening abstracts and were thus not subject to the same scrutiny as the 224 reviewed papers.The 16 studies are listed in Supplementary Table S3.A brief look at the studies suggests that most of them follow the tendency of previous literature in terms of which aspects of DP they investigate.As such, the studies include investigations of holistic processing (Bennetts et al., 2022), of the neural correlates of DP (Parker et al., 2023;Yan et al., 2023;Zhao et al., 2022) andpsychometric investigations (DeGutis et al., 2022;Murray et al., 2022).However, a few of the studies address new questions such as the overlap between DP and autism (Fry et al., 2023;Kamensek et al., 2023) and the role of animal face processing in DP (Epihova et al., 2023).

Future perspectives
Based on the tendencies in the literature outlined in this review, we finally propose that the DP research field could benefit from branching out both methodologically and c o r t e x 1 7 6 ( 2 0 2 4 ) 1 6 1 e1 9 3 theoretically.While there appears to be a vast heterogeneity with regards to the methods used to classify individuals with DP (cf.Fig. 3), the overall methodological approach to DP is predominantly experimental and quantitative.As part of our preregistered protocol, we asked whether each included DP study involved a quantitative or qualitative approach, and this analysis revealed that 210 of the 224 included studies took a quantitative approach, while merely 14 used qualitative or mixed methods designs to investigate DP.As such, future research might benefit from widening its overall methodological scope while narrowing its approach to classifying DP.Further e following the preceding discussion of Fig. 4 e future research might benefit from posing research questions originating outside the fields of cognitive neuropsychology and neuroscience.This could be pursued by studying e.g.how DP is experienced, how it develops throughout life and how it links to other (medical) conditions through the scopes of clinical and developmental psychology.

Concluding comments
The goal of this review was to map the definitions, classification methods and topics of investigation in research on developmental prosopagnosia.This revealed that the DP research field has up till now been characterized by a) a multitude of conceptualizations and definitions of the DP phenomenon, b) a wide range of diagnostic criteria and tools and c) a rather narrow set of research questions mainly borrowed from the field of cognitive neuropsychology and neuroscience.On that basis, we argue that the DP literature is characterized by a conceptual gap, which is on one hand caused by the overwhelming heterogeneity in DP definition and classification and on the other hand by DP's scientific history as a spinoff subject of acquired prosopagnosia.Further we argue that at least one of the reasons why some central and massively posed questions in the DP literature ('is it just faces?'/'is it a deficit in holistic processing?'/'do DP's exhibit a neurotypical N170 potential?')remain unresolved is due to the conceptual gap brought forth here.Given the history of prosopagnosia as a phenomenon originally reported in patients who suffered brain injury (Bodamer, 1947) (Adams et al., 2020;Gru ¨ter et al., 2007;Murray et al., 2018;Yardley et al., 2008).It is, however, rather likely that collection of subjective experience by DP participants might open our eyes to as of yet unanswered (and unasked) questions.One way for the field to progress, then, could be by opening up not only the research questions asked about DP, but also the methods used to investigate them.

Appendix
Table S1

Fig. 1 e
Fig. 1 e Flowchart of the inclusion and exclusion process.Note. Figure layout adopted from Page et al. (2021).
the 224 included studies, roughly half (n ¼ 111) explicitly describe how individuals were classified as having DP.In these 111 studies, a total of 40 different performance assessments, self-report measures and diagnostic interviews are used to classify DP.These include both measures used to assess the degree of face processing difficulties (e.g., the Cambridge Face Memory Test (CFMT,Duchaine & Nakayama  (2006)) and measures used for exclusion (e.g., the Autism Spectrum Quotient (AQ,Baron-Cohen et al. (2001)).Fig.3illustrates how diagnostic measures are combined in studies that explicitly state how they classify individuals with DP (n ¼ 111).While this figure first and foremost conveys the vast diagnostic diversity that dominates the field, an overall diagnostic pattern is clear: DP classification relies in most cases on a combination of an assessment of a) objective performance with memory for unfamiliar faces, b) objective performance with recognition and/or identification of famous faces and c) subjective report of problems with face recognition assessed with questionnaires or interviews.These three types of diagnostic measures are the most frequent in the included records e and has been across the timespan of published papers.There are many variations however, with e.g., no less than 17 different famous faces tests administered across the 111 studies.While this is not surprising e as familiarity tests are highly culturally sensitive e it does constitute a problem for comparing results from DP research across countries and labs.

Fig. 2 e
Fig. 2 e Network graph of definition 'building blocks' of DP.Note.Each node represents a semantic building block used in definitions of DP, with larger nodes indicating building blocks more frequent across the literature.The width of the edges indicates how frequently two building blocks appear together.The plot layout was created with the Fruchterman-Reingold algorithm.

Fig. 4 e
Fig. 4 e Histogram of topics investigated in the DP literature.Note.Infrequent topics (topics that appear only once) were collapsed into one category.
Fig. S1 e Scatter plot of sample size as a function of publication year of all reviewed DP studies.Note.Each dot represents one empirical study.

Fig. S2 e
Fig. S2 e Network graphs of definition 'building blocks' of DP e median split by publication year.Note.Panel A: Definition building blocks in DP studies performed in the years 1991e2014 (n ¼ 98).Panel B: Definition building blocks in DP studies performed in the years 2015e2022 (n ¼ 98).Each node represents a semantic building block used in definitions of DP, with larger nodes indicating building blocks more frequent across the literature.The width of the edges indicates how frequently two building blocks appear together.The plot layout was created with the Fruchterman-Reingold algorithm.

Fig. S3 e
Fig. S3 e Chord diagram of diagnostic tools used to classify individuals with DP e median split by publication year.Note.Panel A: Diagnostic tools used in publications from 2004 to 2013 (n ¼ 24).Panel B: Diagnostic tools used in publications from 2014 to 2022 (n ¼ 87).Each black bar represents one study and the height of the bar indicate the sample size of the DP group of the respective study.A link between two bars indicate that the two tools are used in one study in combination to classify DP.Tools that exist in more than one edition (e.g. the FFT's) are accompanied by references to the study in which that respective edition was used.The distinction between tests of face perception and face memory was done on the basis of the principles outlined by Robotham and Starrfelt (2018).BORB: Birmingham Object Recognition Battery; CFMT: Cambridge Face Memory Test; CFPT: Cambridge Face Perception Test; CCMT: Cambridge Car Memory Test; FFT: Famous Faces Test; FEQ: Faces and Emotions Questionnaire; L-POST: Leuven Perceptual Organization Screening Test; PI20: 20 Item Prosopagnosia Index; VOSP: Visual Object and Space Perception Battery.

Table 1 e
Search string for the literature search in each of the five included databases.
Epihova & Astle, 2024);Kennerknecht, Plu ¨mpe, & Welling, 2008) but it is also possible that both a strictly congenital variant of prosopagnosia and a variant in which the problem occurs as a consequence of environmental influences during childhood exists, and that we are simply unable to diagnostically distinguish between them at present.At least it is clear that for some people with lifelong prosopagnosia, there is a strong hereditary factor(Gru ¨ter et al., 2008;Kennerknecht, Plu ¨mpe, & Welling, 2008), but it is still unclear if this is inherited through different genes than face recognition ability in general, which has a strong genetic component(Wilmer et al., 2010).On this note, we find it remarkable that so relatively few (n ¼ 14) studies investigating DP in children exist (see Supplementary TableS4).While we restricted the scope of this review to studies of DP in the adult population, including the studies investigating children might have helped shed further light on the 'developmental' aspect of the DP concept, and, indeed, more research in how DP develops and manifests in childhood would benefit this endeavor, as others have noted (e.g.,Epihova & Astle, 2024).
Albonico et al., 2017;Bate et al., 2008;¼ 151, 67.4 %) used the term developmental prosopagnosia (DP) to denote the studied phenomenon.Congenital prosopagnosia (CP) is widely used as well (n ¼ 67, 30.0 %), while hereditary prosopagnosia (HPA) is rare (n ¼ 6, 2.7 %).This conceptual dissent in the field warrants an examination of what content is implied when using the terms developmental versus congenital.A congenital disorder is present from birth (following the definitions of CP by e.gAlbonico et al., 2017;Bate et al., 2008; Stollhoff et al.geneticdisposition, complications before or during birth or of environmental influences during childhood.As such, the term DP implies a more broad definition than CP, because it keeps open the possibility that an individual's face processing impairment is not a given certainty from birth.This discrepancy highlights the fact that little is known about how, when and why the disruption of face recognition ability happens.Further research on impaired face recognition in early . BORB: Birmingham Object Recognition Battery; CFMT: Cambridge Face Memory Test; CFPT: Cambridge Face Perception Test; CCMT: Cambridge Car Memory Test; FFT: Famous Faces Test; FEQ: Faces and Emotions Questionnaire; L-POST: Leuven Perceptual Organization Screening Test; PI20: 20 Item Prosopagnosia Index; VOSP: Visual Object and Space Perception Battery.
If DP is indeed a neurodevelopmental phenomenon, we would do well to study it also through lenses of developmental and clinical psychology.It is also curious, that while many definitions include self-report of life long problems with face recognition, the various challenges and impairments reported by participants is rarely the focus of investigation.Indeed, only few studies used more open questionnaires or interview techniques with the aim of gathering information about what is like to have prosopagnosia resting on negative signs (absence of brain injury, absence of general visual impairment), and we suggest the field would benefit from future research that seeks to fill its conceptual gap by also investigating what DP is, rather than what it is not.To this end, more research is needed on how DP develops, how it runs in families and how it progresses through life.In that endeavor, the DP field could benefit from branching out to publication outlets outside of cognitive neuropsychology and neuroscience.

e
Characteristics of the 224 included studies of DP.

Table S2 e
List of infrequent topics in the DP literature

Table S4 e
List of studies investigating DP in children.