Abstract
Recently some epistemologists have approached the question whether epistemic justification comes in degrees from a linguistic perspective. Drawing insights from linguistic analyses of gradable adjectives, they investigate whether epistemic occurrences of ‘justified’ are gradable and if yes what type of gradability they involve. These authors conclude that the adjective passes standard tests for gradability, but they classify it as belonging to different categories: as either an absolute or a relative gradable adjective. The aim of this paper is to further clarify the question of what kind of gradability is instantiated by epistemic uses of ‘justified’, and to investigate the consequences that this may have on epistemological theorizing. In particular, we challenge the alleged evidence provided by Siscoe (2021: 517–59) for the relative gradability of ‘justified’ and we provide positive arguments for the claim that ‘justified’ is a specific kind of absolute gradable adjective (a totally closed scale absolute with default minimal reading). We consider some important philosophical implications of these results. We also argue that the scale type of ‘justified’ is ordinal, and we consider which other properties typically associated to justification possess a similar type of scale. We argue that several popular contemporary models of graded justification, such as probabilistic and reliabilist accounts and various kinds of Lockean views, do not fit well with our results about the gradability of ‘justified’. Linguistic data rather favour normalcy and plausibility accounts of graded justification.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Staffel (2019: §8.4) classifies ‘rational’ as an absolute gradable, but she also treats justification and rationality as equivalent or interchangeable concepts (see e.g., 2019: 81–82, ch.8 and 9).
For instance Aidan McGlynn claims that “it is relatively uncontroversial that justification comes in degrees, but knowledge does not. How does this account [that one’s belief is justified if and only if it conforms with the knowledge norm of belief] explain how one belief can be more justified that another, if justified belief just is knowledge, and whether one knows isn’t a gradable matter?” (2014: 40–41). See also Smith 2016: 95. Sutton (2007: Ch2, fn4) explicitly recognizes that on a view equating knowledge and justification, comparisons of justified beliefs are strictly speaking nonsense.
We will provide an overview of different types of gradable adjectives in the next section.
In what follows we leave open the possibility that absolute gradable adjectives are gradable only as long as they are coerced into graded uses (due to loose talk), and thus are only derivatively gradable. See Burnett (2017) for a theory elaborating this option for absolute adjectives in general, and Hawthorne & Logins (2021) and Logins (forthcoming) for similar suggestions with respect to ‘justified’.
While most relatives lie on fully open scales (scales without any upper or lower boundaries), there are alleged cases of relatives taking bounded scales (e.g., cheap, slow, quiet, probable). Lassiter (2017): §4.2.
The above classification involves some important simplifications. A more standard classification in scalar semantics involves two orthogonal distinctions: the first one concerns features of the adjective’s scale, which could be totally open, lower closed, upper closed, or totally closed. The second distinction concerns the nature of the standard of comparison, whether this is fixed contextually (relative) or by some non-contextual (minimal or maximal) threshold (absolute). However, all adjectives with totally open scales are relatives; absolute adjectives with upper open scales are all minimals, and those with lower open scales are all maximals. Fully closed scale adjectives typically allow both maximal and minimal readings, but tend to default in one of them. These limited combinations determine the categories of absolute gradable adjectives mentioned in the main text. While less precise, our classification is simpler and more convenient given our purposes. The reader could easily map our categories into the more familiar classification.
Kennedy and McNally (2005: 361). Notice however that non-default readings of totals are always possible given specific features of the context. For instance, the adjective ‘open’ has a default minimal reading, but it can get a maximal interpretation in some contexts, such as when it is combined with maximal modifiers such as ‘completely’. See Kennedy 2007: §4.3. We will come back to this point in §§5–6.
In what follows we abstract from specific differences in the formal semantics developed by these linguists and focus on the empirical data and the tests upon which there is consensus. In what follows we adopt the terminology from Kennedy and McNally, who in turn take it from Unger (1975).
Cf. Hawthorne & Logins (2021). The reason why absolutes imply their antonyms but relatives don’t is pretty straightforward: absolutes pick either minimal or maximal degrees on a scale. A minimal positive degree corresponds to a maximal negative degree on the same scale. For instance the minimal degree of wetness/impurity corresponds to the maximum degree of dryness/purity. This doesn’t apply to relatives, whose standards (both positive and negative) are contextually identified and allow grey areas. See Kennedy and McNally (2005: 359-360).
As a piece of anecdotal evidence, a poll on Twitter with 68 participants showed that only 32.4% of them found ‘very justified’ fine, while 67.6% found ‘very justified’ marginal or pertaining to a creative use of language. Some data from English corpuses: in the Corpus of Contemporary English (COCA), ‘completely justified’ has 95 occurrences;’fully justified’ 114 occurrences; ‘very justified’ only 14 occurrences.
The above remarks are fully compatible with non-maximal uses of ‘completely justified’. Indeed, if ‘justified’ is a total with default minimal readings, it may be common to find cases in which ‘completely’ and ‘totally’ modify a minimal use of the adjective. In such cases, the modifier may express emphasis or high speaker confidence. Consider how ‘completely’ modifies the total absolute ‘open’: depending on the context, ‘completely open’ could mean ‘maximally open’ (maximization), ‘very much open’ (emphasis), or ‘there is no doubt that it’s open’ (high confidence).
Probabilistic reasoning is known to be a highly cognitively demanding task triggering heuristics and often affected by biases (cf. Kahneman and Tversky 1981).
A reviewer rightly observes that, while the results of Siscoe’s study do not support his preferred view, these results are difficult to explain also for other views, including those defended in this paper. The reviewer is correct. In fact we don’t think that Siscoe’s results could be explained by any interpretation of ‘justified’, be it relative or absolute. This further supports our contention that the tests’ results are unreliable. Indeed, we repeated Siscoe’s test with a larger sample of participants, but we couldn’t replicate his results. An alternative explanation of such results could be that the question “If I asked you to point at the one who is justified […], who would you point to?” involves a presupposition that one of the subjects is justified. This question could be interpreted as follows: “Assuming that one of the subjects is justified in his belief, who would you point to?”. This interpretation would draw the respondent’s attention away from the chance’s absolute values and force her to make a choice between the two subjects.
The same point applies to ‘rational’, which Siscoe classifies as an absolute gradable adjective (see also Staffel 2020: §8.4). Intuitively, we cannot differentiate between Joe and Frank with a use of (N’):
(N’) #Point to the one who is rational in believing that the Shooters won.
Similar examples involving tolerant contexts can be constructed for minimal adjectives. Consider the minimal adjective ‘poisonous’. In a strict sense, any small degree of poison is sufficient to make a drink poisonous. However, when we use this adjective in most ordinary contexts we typically allow tolerant uses involving a certain degree of imprecision (Burnett 2014: 7). For instance, when we consider whether a drink would be sufficiently poisonous to cause serious health effects, we use ‘poisonous’ in a non-strict sense, such that adding or subtracting some particles of the poisonous substance would not make a difference in whether we would call it poisonous. In such a context, ± 1 mg is an indifference relation for ‘poisonous’.
See also Pinkal (1995). Notice that the current point is not that precisification is impossible for relatives, only that such a phenomenon is less natural for such adjectives due to the intrinsic context-sensitivity of relative standards. By contrast, since for absolutes the maximal or minimal standards are fixed by the meaning of the term, it is natural to specify the degree of precision with which one refers to such standards.
Siscoe doesn’t specify whether and how weak the support provided by Joe’s second-hand testimony should be considered. He seems to take for granted that second-hand evidence is worse than other types of evidence such as perception. This assumption is unwarranted. Suppose for instance that Frank has a very bad memory and Joe’s second-hand knowledge is based on the testimony of the Shooters’ coach. With such assumptions in place, (16) would sound plainly inappropriate. For the sake of argument we assume that in Siscoe’s case Frank’s first-hand knowledge is more robust than Joe’s second-hand one.
This is easily explainable by the fact that minimals have fixed minimal endpoints. If something definitely meets the minimal standard for wetness or openness, we cannot precisify the context in such a way that that thing doesn’t count as wet or open.
Compare to Smith (2016: 97).
The study was conducted on Amazon Mechanical Turk, using the MT Masters function.
Alternatively, we could imagine a situation in which a subject S has slightly more evidential support for p than a different subject S*, but the support is not sufficient to fully justify either subject. Still, it could make sense to say that S is more justified to believe that p than S* is, though neither subject is justified.
Notice that in some special cases also absolute adjectives can be coerced into relative interpretations. See Kennedy and McNally (2005): 370–1. In such special contexts, absolute adjectives such as ‘justified’ could figure in comparative judgments in which no entailment is detected. Such cases are however quite uncommon.
From the Corpus of Contemporary English (COCA). Maximal modifiers: ‘fully justified’ 114 occurrences; ‘completely justified’: 95 occurrences; ‘perfectly justified’: 71; ‘totally justified’: 56; ‘absolutely justified’ 19; ‘almost justified’: 5 (note: ‘justified’ used as adjective only in 1 occurrence); ‘100% justified’: 3; ‘maximally justified’: 0. Minimal modifiers: ‘somewhat justified’: 36; ‘partly/partially justified’: 27; ‘quite justified’ 22; ‘slightly justified’: 1; ‘minimally justified’: 0. The above data suggest that the degree modifier ‘slightly’ doesn’t combine well with ‘justified’. This may seem surprising, since many linguists take this modifier to be one of the key identifiers for minimal absolutes. However, a more careful look at the distribution of this modifier shows that ‘slighlty’ doesn’t combine well with most gradable minimal and total deverbal adjectives (adjectives derived from verbs). Examples of minimal deverbal adjectives are ‘deserved’, ‘criticized’ and ‘needed’; examples of totals include ‘written’, ‘acquainted’, ‘documented’, ‘educated’ and ‘justified’. ‘Slightly’ also doesn’t combine well with other non-deverbal minimal adjectives such as ‘impure’ and ‘uncertain’. An investigation into the reasons of such exceptions goes beyond the scope of the present paper.
Kennedy and McNally (2005), §6.
The example is adapted from Kennedy 2007: 38.
For a certainty based account see Beddor (2020). For knowledge-based theories of justification see Bird (2007), Littlejohn (2013) and Ichikawa (2014). In particular, Bird suggests an account of degrees of justification based on how close beliefs come to satisfying the mental requirements for knowledge and how safely they meet these requirements.
See also Schulz (2022).
Ranking theorists use the symbol of infinite to represent the highest rank, but as Smith (2016: 157, fn2) observes, in a ranking function the number of ranks is not necessarily meant to be infinite.
It is common to assume that ranking functions are regular: for any non-impossible proposition p, r(p) < ∞ (Spohn 2009: §2.3; Smith 2016: 98). With this assumption in place only necessary propositions have infinite rank and could count as completely justified. This seems to clash with the ordinary intuition that one could have a maximal degree of justification for believing contingent propositions (e.g., that I am not in Dushanbe right now). In order to avoid this consequence, one could reject regularity and allow that also for some contingent propositions p, r(p) = ∞. Since a ranking function is defined over a space of possibilities (or a set of possible worlds), we can do this by restricting the domain of relevant possibilities so that certain contingent propositions count as true in every relevant possible world. A way to do this is by interpreting ranking functions as representing degrees of justification relative to a body of evidence. Given such interpretation, propositions that are entailed by the evidence will also have maximal justification. Relatedly, there will be contingent propositions that hold throughout all of the possible worlds under consideration, since we will only be considering worlds in which the evidence holds. Thanks to an anonymous reviewer for helpful comments on this point.
Ranking theory is just one of the formal frameworks consistent with a total interpretation of ‘justified’. Others could fit the purpose as well.
There may be some doubt whether ‘reliable’ is a relative. While this adjective passes most tests for relatives, sometimes it can combine with absolute modifiers.
Thanks to an anonymous reviewer for encouraging us to make this assumption explicit and address the related potential worry. It is important to stress that not all arguments in this section rely on this assumption. For instance, our considerations about Lockean views and ranking function models do not appeal to structural similarities of the scalar features of justification and its grounding properties. They rather rely on general considerations about the structure of the justification scale, which are independent of what properties ground such a scale. Other arguments concern the dimension of the justification’s scale. Assuming that gradable properties are partially constituted by scales over which such properties’ degrees are measured (see §2), in these arguments we consider what the dimension of the scale constitutive of justification could be (probability, safety…). Importantly, on this view gradable properties are not merely grounded in a scalar dimension, but constituted by it. If so, gradable features of the scale’s dimension must be structurally analogous with those of the properties measured on such scale.
In the Corpus of Contemporary English (COCA), ‘half justified’ has only 1 occurrence;’x% justified’ only 2 occurrences (both occurrences are ‘100% justified’).
The test of acceptability of proportional modifiers doesn’t apply to ‘safe’ and ‘dangerous’ because their scales are not fully bounded. Since these adjectives’ scales do not have a minimum or a maximum endpoint, expressions such as ‘X is half/n% safe/dangerous’ are undefined. See Lassiter 2017: §2.5.2 for further arguments that danger and safety have interval scales.
References
Beddor, B. (2020). New work for certainty. Philosophers' Imprint 20 (8).
Bird, A. (2007). Justified judging. Philosophy and Phenomenological Research, 74(1), 81–110.
Burnett, H. (2014). ‘A delineation solution to the puzzles of absolute adjectives. Linguistics and Philosophy, 37(1), 1–39.
Burnett, H. (2017). Gradability in natural language: Logical and grammatical foundations. Oxford University Press.
Clarke, R. (2013). Belief is credence one in context. Philosopher’s Imprint, 13, 1–18.
Cohen, S. (1998). Contextualist solutions to epistemological problems: skepticism, Gettier, and the lottery. Australasian Journal of Philosophy, 76, 289–306.
Cresswell, M. J. (1976). The semantics of degree. In B. Partee (Ed.), Montague Grammar (pp. 261–292). Academic Press.
Cruse, D. A. (1986). Lexical Semantics. Cambridge University Press.
Dorst, K. (2019). Lockeans maximize expected accuracy. Mind, 128(509), 175–211.
Dretske, F. (1970). Epistemic operators. Journal of Philosophy, 67, 1007–1023.
Foley, R. (2009). Beliefs, Degrees of Belief, and the Lockean Thesis. In F. Huber & C. Schmidt-Petri (Eds.), Degrees of Belief (pp. 37–47). Springer.
Greco, D. (2015). How i learned to stop worrying and love probability 1. Philosophical Perspectives, 29, 179–201.
Hawthorne, J., & Logins, A. (2021). Graded epistemic justification. Philosophical Studies, 178(6), 1845–1858.
Ichikawa, J. (2014). Justification is potential knowledge. Canadian Journal of Philosophy, 44(2), 184–206.
Kahneman, D., & Tversky, A. (1981). The framing of decisions and the psychology of choice. Science, 211(4481), 453–458.
Kennedy, C. (1999). Projecting the Adjective: The Syntax and Semantics of Gradability and Comparison. Outstanding Dissertations in Linguistics Series. Garland Press.
Kennedy, C. (2007). Vagueness and grammar: the semantics of relative and absolute gradable adjectives. Linguistics and Philosophy, 30(1), 1–45.
Kennedy, C., & McNally, L. (2005). Scale structure and the semantic typology of gradable predicates. Language, 81(2), 345–381.
Lassiter, D. (2010). Gradable epistemic modals, probability, and scale structure. In N. Li & D. Lutz (eds.), Semantics and Linguistic Theory (SALT) 20. pp. 197–215.
Lassiter, D. (2017). Graded Modality: Qualitative and Quantitative Perspectives. Oxford University Press.
Leitgeb, H. (2014). The stability theory of belief. Philosophical Review, 123(2), 131–171.
Levi, I. (1991). The Fixation of Belief and its Undoing: Changing Beliefs Through Inquiry. CUP.
Littlejohn, C. (2013). The russellian retreat. Proceedings of the Aristotelian Society 113(3pt3) pp. 293–320.
Logins, A. (2022). Is an increase in probability always an increase in evidential support? Erkenntnis, 87, 1231–1255. https://doi.org/10.1007/s10670-020-00241-4
Logins, A. (forthcoming). The paradox of graded justification. Episteme. https://doi.org/10.1017/epi.2022.51
McGlynn, A. (2014). Knowledge First? Palgrave Macmillan.
Pinkal, M. (1995). Logic and Lexicon. Kluwer.
Rotstein, C., & Winter, Y. (2004). Total adjectives vs. partial adjectives: scale structure and higher-order modifiers. Natural Language Semantics, 12, 259–288.
Schulz, M. (2022). Degrees of doxastic justification. Erkenntnis, 87(6), 2943–2972.
Siegel, M. (1979). Measure adjectives in Montague grammar. In S. Davis & M. Mithun (Eds.), Linguistics, philosophy, and Montague grammar (pp. 223–262). University Press of Texas.
Siscoe, R. W. (2021). Belief, rational and justified. Mind, 130(517), 59–83.
Siscoe, R. W. (2022). Real and ideal rationality. Philosophical Studies, 179(3), 879–910.
Smith, M. (2016). Between Probability and Certainty: What Justifies Belief. Oxford University Press.
Spohn, W. (2009). A survey of ranking theory. in Franz Huber & Christoph Schmidt-petri (eds.), Degrees of Belief. Springer.
Spohn, W. (2012). The Laws of Belief: Ranking Theory and its Philosophical Applications. Oxford University Press.
Staffel, J. (2019). Unsettled Thoughts: A Theory of Degrees of Rationality. Oxford University Press.
Sutton, J. (2007). Without Justification. MIT Press.
Unger, P. (1975). Ignorance. Clarendon Press.
Williamson, T. (2006). Must do better. Truth and Realism. Oxford University Press. pp. 177–187.
Williamson, T. (forthcoming). Knowledge, credence, and strength of belief, in A.K. Flowerree and Baron Reed (eds.), Expansive Epistemology: Norms, Action, and the Social World. Routledge.
Acknowledgments
We would like to thank two anonymous reviewers for helpful comments on earlier drafts of this article. Earlier versions of this article were presented at the Lingnan Philosophy Seminar, at the Frontiers of Epistemic Agency workshop at Wuhan University, and at the Zurich XPhi Lab November 2021 meeting at the University of Zurich. Thanks to the audiences for their helpful feedback. Thanks to Kevin Reuter for a discussion on an earlier version of this article. Part of the research work that led to this article was supported by the Swiss National Science Foundation Grant ‘Graded Epistemology’ Number 186137.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Fassio, D., Logins, A. Justification and gradability. Philos Stud 180, 2051–2077 (2023). https://doi.org/10.1007/s11098-023-01945-3
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11098-023-01945-3